)]}' { "log": [ { "commit": "92d15c2ccbb3e31a3fc71ad28fdb55e1319383c0", "tree": "8d83c0dc3c6b935d8367e331872f242b742f0a8a", "parents": [ "f20bf6125605acbbc7eb8c9420d7221c91aa83eb", "644bd2f048972d75eb1979b1fdca257d528ce687" ], "author": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 10:09:16 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 10:09:16 2007 -0700" }, "message": "Merge branch \u0027for-linus\u0027 of git://git.kernel.dk/data/git/linux-2.6-block\n\n* \u0027for-linus\u0027 of git://git.kernel.dk/data/git/linux-2.6-block: (63 commits)\n Fix memory leak in dm-crypt\n SPARC64: sg chaining support\n SPARC: sg chaining support\n PPC: sg chaining support\n PS3: sg chaining support\n IA64: sg chaining support\n x86-64: enable sg chaining\n x86-64: update pci-gart iommu to sg helpers\n x86-64: update nommu to sg helpers\n x86-64: update calgary iommu to sg helpers\n swiotlb: sg chaining support\n i386: enable sg chaining\n i386 dma_map_sg: convert to using sg helpers\n mmc: need to zero sglist on init\n Panic in blk_rq_map_sg() from CCISS driver\n remove sglist_len\n remove blk_queue_max_phys_segments in libata\n revert sg segment size ifdefs\n Fixup u14-34f ENABLE_SG_CHAINING\n qla1280: enable use_sg_chaining option\n ...\n" }, { "commit": "e2fc88d0643ca68f2011e6db4aa31e22bd94210c", "tree": "0865518013da9b0209a3c1c55111b46cb789b712", "parents": [ "dbcb0f19c877df9026b8c1227758d38bd561e9c4" ], "author": { "name": "Adrian Bunk", "email": "bunk@stusta.de", "time": "Tue Oct 16 01:26:27 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:03 2007 -0700" }, "message": "mm/vmstat.c: cleanups\n\nThis patch contains the following cleanups:\n- make the needlessly global setup_vmstat() static\n- remove the unused refresh_vm_stats()\n\nSigned-off-by: Adrian Bunk \u003cbunk@stusta.de\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "dbcb0f19c877df9026b8c1227758d38bd561e9c4", "tree": "f58c85976906f42ff44798f514177392d7c48d0f", "parents": [ "d8dc74f212c38407fc9f4367181f8f969b719485" ], "author": { "name": "Adrian Bunk", "email": "bunk@stusta.de", "time": "Tue Oct 16 01:26:26 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:03 2007 -0700" }, "message": "mm/mempolicy.c: cleanups\n\nThis patch contains the following cleanups:\n- every file should include the headers containing the prototypes for\n its global functions\n- make the follosing needlessly global functions static:\n - migrate_to_node()\n - do_mbind()\n - sp_alloc()\n - mpol_rebind_policy()\n\n[akpm@linux-foundation.org: fix uninitialised var warning]\nSigned-off-by: Adrian Bunk \u003cbunk@stusta.de\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "d8dc74f212c38407fc9f4367181f8f969b719485", "tree": "2ff1ec0aaee1e6b3ee29d258a7158deb4b43688c", "parents": [ "af767cbdd78f293485c294113885d95e7f1da123" ], "author": { "name": "Adrian Bunk", "email": "bunk@stusta.de", "time": "Tue Oct 16 01:26:26 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:03 2007 -0700" }, "message": "mm/shmem.c: make 3 functions static\n\nThis patch makes three needlessly global functions static.\n\nSigned-off-by: Adrian Bunk \u003cbunk@stusta.de\u003e\nCc: Hugh Dickins \u003chugh@veritas.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "af767cbdd78f293485c294113885d95e7f1da123", "tree": "934d5c6180e808ee2062d8d2ef2ed758c452286d", "parents": [ "63b4613c3f0d4b724ba259dc6c201bb68b884e1a" ], "author": { "name": "Adam Litke", "email": "agl@us.ibm.com", "time": "Tue Oct 16 01:26:25 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:03 2007 -0700" }, "message": "hugetlb: fix dynamic pool resize failure case\n\nWhen gather_surplus_pages() fails to allocate enough huge pages to satisfy\nthe requested reservation, it frees what it did allocate back to the buddy\nallocator. put_page() should be called instead of update_and_free_page()\nto ensure that pool counters are updated as appropriate and the page\u0027s\nrefcount is decremented.\n\nSigned-off-by: Adam Litke \u003cagl@us.ibm.com\u003e\nAcked-by: Dave Hansen \u003chaveblue@us.ibm.com\u003e\nCc: David Gibson \u003chermes@gibson.dropbear.id.au\u003e\nCc: William Lee Irwin III \u003cwli@holomorphy.com\u003e\nCc: Badari Pulavarty \u003cpbadari@us.ibm.com\u003e\nCc: Ken Chen \u003ckenchen@google.com\u003e\nCc: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "63b4613c3f0d4b724ba259dc6c201bb68b884e1a", "tree": "878818a47052fd204aa0a5d34e592967732d59f9", "parents": [ "6b0c880dfefecedb9ad353014ed41505c32aca82" ], "author": { "name": "Nishanth Aravamudan", "email": "nacc@us.ibm.com", "time": "Tue Oct 16 01:26:24 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:03 2007 -0700" }, "message": "hugetlb: fix hugepage allocation with memoryless nodes\n\nAnton found a problem with the hugetlb pool allocation when some nodes have\nno memory (http://marc.info/?l\u003dlinux-mm\u0026m\u003d118133042025995\u0026w\u003d2). Lee worked\non versions that tried to fix it, but none were accepted. Christoph has\ncreated a set of patches which allow for GFP_THISNODE allocations to fail\nif the node has no memory.\n\nCurrently, alloc_fresh_huge_page() returns NULL when it is not able to\nallocate a huge page on the current node, as specified by its custom\ninterleave variable. The callers of this function, though, assume that a\nfailure in alloc_fresh_huge_page() indicates no hugepages can be allocated\non the system period. This might not be the case, for instance, if we have\nan uneven NUMA system, and we happen to try to allocate a hugepage on a\nnode with less memory and fail, while there is still plenty of free memory\non the other nodes.\n\nTo correct this, make alloc_fresh_huge_page() search through all online\nnodes before deciding no hugepages can be allocated. Add a helper function\nfor actually allocating the hugepage. Use a new global nid iterator to\ncontrol which nid to allocate on.\n\nNote: we expect particular semantics for __GFP_THISNODE, which are now\nenforced even for memoryless nodes. That is, there is should be no\nfallback to other nodes. Therefore, we rely on the nid passed into\nalloc_pages_node() to be the nid the page comes from. If this is\nincorrect, accounting will break.\n\nTested on x86 !NUMA, x86 NUMA, x86_64 NUMA and ppc64 NUMA (with 2\nmemoryless nodes).\n\nBefore on the ppc64 box:\nTrying to clear the hugetlb pool\nDone. 0 free\nTrying to resize the pool to 100\nNode 0 HugePages_Free: 25\nNode 1 HugePages_Free: 75\nNode 2 HugePages_Free: 0\nNode 3 HugePages_Free: 0\nDone. Initially 100 free\nTrying to resize the pool to 200\nNode 0 HugePages_Free: 50\nNode 1 HugePages_Free: 150\nNode 2 HugePages_Free: 0\nNode 3 HugePages_Free: 0\nDone. 200 free\n\nAfter:\nTrying to clear the hugetlb pool\nDone. 0 free\nTrying to resize the pool to 100\nNode 0 HugePages_Free: 50\nNode 1 HugePages_Free: 50\nNode 2 HugePages_Free: 0\nNode 3 HugePages_Free: 0\nDone. Initially 100 free\nTrying to resize the pool to 200\nNode 0 HugePages_Free: 100\nNode 1 HugePages_Free: 100\nNode 2 HugePages_Free: 0\nNode 3 HugePages_Free: 0\nDone. 200 free\n\nSigned-off-by: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: Adam Litke \u003cagl@us.ibm.com\u003e\nCc: David Gibson \u003chermes@gibson.dropbear.id.au\u003e\nCc: Badari Pulavarty \u003cpbadari@us.ibm.com\u003e\nCc: Ken Chen \u003ckenchen@google.com\u003e\nCc: William Lee Irwin III \u003cwli@holomorphy.com\u003e\nCc: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6b0c880dfefecedb9ad353014ed41505c32aca82", "tree": "2cf472a539f6de265deede9fb580be821bf9dda1", "parents": [ "e63e1e5a6b75416ab758025fd80e247c4f103b1b" ], "author": { "name": "Adam Litke", "email": "agl@us.ibm.com", "time": "Tue Oct 16 01:26:23 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:03 2007 -0700" }, "message": "hugetlb: fix pool resizing corner case\n\nWhen shrinking the size of the hugetlb pool via the nr_hugepages sysctl, we\nare careful to keep enough pages around to satisfy reservations. But the\ncalculation is flawed for the following scenario:\n\nAction Pool Counters (Total, Free, Resv)\n\u003d\u003d\u003d\u003d\u003d\u003d \u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nSet pool to 1 page 1 1 0\nMap 1 page MAP_PRIVATE 1 1 0\nTouch the page to fault it in 1 0 0\nSet pool to 3 pages 3 2 0\nMap 2 pages MAP_SHARED 3 2 2\nSet pool to 2 pages 2 1 2 \u003c-- Mistake, should be 3 2 2\nTouch the 2 shared pages 2 0 1 \u003c-- Program crashes here\n\nThe last touch above will terminate the process due to lack of huge pages.\n\nThis patch corrects the calculation so that it factors in pages being used\nfor private mappings. Andrew, this is a standalone fix suitable for\nmainline. It is also now corrected in my latest dynamic pool resizing\npatchset which I will send out soon.\n\nSigned-off-by: Adam Litke \u003cagl@us.ibm.com\u003e\nAcked-by: Ken Chen \u003ckenchen@google.com\u003e\nCc: David Gibson \u003cdavid@gibson.dropbear.id.au\u003e\nCc: Badari Pulavarty \u003cpbadari@us.ibm.com\u003e\nCc: William Lee Irwin III \u003cwli@holomorphy.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "54f9f80d6543fb7b157d3b11e2e7911dc1379790", "tree": "70d34e45f4d1694609db25da0ccad90b4b0ad473", "parents": [ "e4e574b767ba63101cfda2b42d72f38546319297" ], "author": { "name": "Adam Litke", "email": "agl@us.ibm.com", "time": "Tue Oct 16 01:26:20 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:02 2007 -0700" }, "message": "hugetlb: Add hugetlb_dynamic_pool sysctl\n\nThe maximum size of the huge page pool can be controlled using the overall\nsize of the hugetlb filesystem (via its \u0027size\u0027 mount option). However in the\ncommon case the this will not be set as the pool is traditionally fixed in\nsize at boot time. In order to maintain the expected semantics, we need to\nprevent the pool expanding by default.\n\nThis patch introduces a new sysctl controlling dynamic pool resizing. When\nthis is enabled the pool will expand beyond its base size up to the size of\nthe hugetlb filesystem. It is disabled by default.\n\nSigned-off-by: Adam Litke \u003cagl@us.ibm.com\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Dave McCracken \u003cdave.mccracken@oracle.com\u003e\nCc: William Irwin \u003cbill.irwin@oracle.com\u003e\nCc: David Gibson \u003cdavid@gibson.dropbear.id.au\u003e\nCc: Ken Chen \u003ckenchen@google.com\u003e\nCc: Badari Pulavarty \u003cpbadari@us.ibm.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "e4e574b767ba63101cfda2b42d72f38546319297", "tree": "084b94d01c71ccd898f8df0ec441e6726e657e75", "parents": [ "7893d1d505d59db9d4f35165c8b6d3c6dff40a32" ], "author": { "name": "Adam Litke", "email": "agl@us.ibm.com", "time": "Tue Oct 16 01:26:19 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:02 2007 -0700" }, "message": "hugetlb: Try to grow hugetlb pool for MAP_SHARED mappings\n\nShared mappings require special handling because the huge pages needed to\nfully populate the VMA must be reserved at mmap time. If not enough pages are\navailable when making the reservation, allocate all of the shortfall at once\nfrom the buddy allocator and add the pages directly to the hugetlb pool. If\nthey cannot be allocated, then fail the mapping. The page surplus is\naccounted for in the same way as for private mappings; faulted surplus pages\nwill be freed at unmap time. Reserved, surplus pages that have not been used\nmust be freed separately when their reservation has been released.\n\nSigned-off-by: Adam Litke \u003cagl@us.ibm.com\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Dave McCracken \u003cdave.mccracken@oracle.com\u003e\nCc: William Irwin \u003cbill.irwin@oracle.com\u003e\nCc: David Gibson \u003cdavid@gibson.dropbear.id.au\u003e\nCc: Ken Chen \u003ckenchen@google.com\u003e\nCc: Badari Pulavarty \u003cpbadari@us.ibm.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "7893d1d505d59db9d4f35165c8b6d3c6dff40a32", "tree": "6bea3b41e111b1d1774980296a032012a3926e9c", "parents": [ "6af2acb6619688046039234f716fd003e6ed2b3f" ], "author": { "name": "Adam Litke", "email": "agl@us.ibm.com", "time": "Tue Oct 16 01:26:18 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:02 2007 -0700" }, "message": "hugetlb: Try to grow hugetlb pool for MAP_PRIVATE mappings\n\nBecause we overcommit hugepages for MAP_PRIVATE mappings, it is possible that\nthe hugetlb pool will be exhausted or completely reserved when a hugepage is\nneeded to satisfy a page fault. Before killing the process in this situation,\ntry to allocate a hugepage directly from the buddy allocator.\n\nThe explicitly configured pool size becomes a low watermark. When dynamically\ngrown, the allocated huge pages are accounted as a surplus over the watermark.\n As huge pages are freed on a node, surplus pages are released to the buddy\nallocator so that the pool will shrink back to the watermark.\n\nSurplus accounting also allows for friendlier explicit pool resizing. When\nshrinking a pool that is fully in-use, increase the surplus so pages will be\nreturned to the buddy allocator as soon as they are freed. When growing a\npool that has a surplus, consume the surplus first and then allocate new\npages.\n\nSigned-off-by: Adam Litke \u003cagl@us.ibm.com\u003e\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Dave McCracken \u003cdave.mccracken@oracle.com\u003e\nCc: William Irwin \u003cbill.irwin@oracle.com\u003e\nCc: David Gibson \u003cdavid@gibson.dropbear.id.au\u003e\nCc: Ken Chen \u003ckenchen@google.com\u003e\nCc: Badari Pulavarty \u003cpbadari@us.ibm.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6af2acb6619688046039234f716fd003e6ed2b3f", "tree": "6afd273778dcbc4b2706a793c756a1ccd00a44f3", "parents": [ "98f3cfc1dc7a53b629d43b7844a9b3f786213048" ], "author": { "name": "Adam Litke", "email": "agl@us.ibm.com", "time": "Tue Oct 16 01:26:16 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:02 2007 -0700" }, "message": "hugetlb: Move update_and_free_page\n\nDynamic huge page pool resizing.\n\nIn most real-world scenarios, configuring the size of the hugetlb pool\ncorrectly is a difficult task. If too few pages are allocated to the pool,\napplications using MAP_SHARED may fail to mmap() a hugepage region and\napplications using MAP_PRIVATE may receive SIGBUS. Isolating too much memory\nin the hugetlb pool means it is not available for other uses, especially those\nprograms not using huge pages.\n\nThe obvious answer is to let the hugetlb pool grow and shrink in response to\nthe runtime demand for huge pages. The work Mel Gorman has been doing to\nestablish a memory zone for movable memory allocations makes dynamically\nresizing the hugetlb pool reliable within the limits of that zone. This patch\nseries implements dynamic pool resizing for private and shared mappings while\nbeing careful to maintain existing semantics. Please reply with your comments\nand feedback; even just to say whether it would be a useful feature to you.\nThanks.\n\nHow it works\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n\nUpon depletion of the hugetlb pool, rather than reporting an error immediately,\nfirst try and allocate the needed huge pages directly from the buddy allocator.\nCare must be taken to avoid unbounded growth of the hugetlb pool, so the\nhugetlb filesystem quota is used to limit overall pool size.\n\nThe real work begins when we decide there is a shortage of huge pages. What\nhappens next depends on whether the pages are for a private or shared mapping.\nPrivate mappings are straightforward. At fault time, if alloc_huge_page()\nfails, we allocate a page from the buddy allocator and increment the source\nnode\u0027s surplus_huge_pages counter. When free_huge_page() is called for a page\non a node with a surplus, the page is freed directly to the buddy allocator\ninstead of the hugetlb pool.\n\nBecause shared mappings require all of the pages to be reserved up front, some\nadditional work must be done at mmap() to support them. We determine the\nreservation shortage and allocate the required number of pages all at once.\nThese pages are then added to the hugetlb pool and marked reserved. Where that\nis not possible the mmap() will fail. As with private mappings, the\nappropriate surplus counters are updated. Since reserved huge pages won\u0027t\nnecessarily be used by the process, we can\u0027t be sure that free_huge_page() will\nalways be called to return surplus pages to the buddy allocator. To prevent\nthe huge page pool from bloating, we must free unused surplus pages when their\nreservation has ended.\n\nControlling it\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n\nWith the entire patch series applied, pool resizing is off by default so unless\nspecific action is taken, the semantics are unchanged.\n\nTo take advantage of the flexibility afforded by this patch series one must\ntolerate a change in semantics. To control hugetlb pool growth, the following\ntechniques can be employed:\n\n * A sysctl tunable to enable/disable the feature entirely\n * The size\u003d mount option for hugetlbfs filesystems to limit pool size\n\nPerformance\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n\nWhen contiguous memory is readily available, it is expected that the cost of\ndynamicly resizing the pool will be small. This series has been performance\ntested with \u0027stream\u0027 to measure this cost.\n\nStream (http://www.cs.virginia.edu/stream/) was linked with libhugetlbfs to\nenable remapping of the text and data/bss segments into huge pages.\n\nStream with small array\n-----------------------\nBaseline: \tnr_hugepages \u003d 0, No libhugetlbfs segment remapping\nPreallocated:\tnr_hugepages \u003d 5, Text and data/bss remapping\nDynamic:\tnr_hugepages \u003d 0, Text and data/bss remapping\n\n\t\t\t\tRate (MB/s)\nFunction\tBaseline\tPreallocated\tDynamic\nCopy:\t\t4695.6266\t5942.8371\t5982.2287\nScale:\t\t4451.5776\t5017.1419\t5658.7843\nAdd:\t\t5815.8849\t7927.7827\t8119.3552\nTriad:\t\t5949.4144\t8527.6492\t8110.6903\n\nStream with large array\n-----------------------\nBaseline: \tnr_hugepages \u003d 0, No libhugetlbfs segment remapping\nPreallocated:\tnr_hugepages \u003d 67, Text and data/bss remapping\nDynamic:\tnr_hugepages \u003d 0, Text and data/bss remapping\n\n\t\t\t\tRate (MB/s)\nFunction\tBaseline\tPreallocated\tDynamic\nCopy:\t\t2227.8281\t2544.2732\t2546.4947\nScale:\t\t2136.3208\t2430.7294\t2421.2074\nAdd:\t\t2773.1449\t4004.0021\t3999.4331\nTriad:\t\t2748.4502\t3777.0109\t3773.4970\n\n* All numbers are averages taken from 10 consecutive runs with a maximum\n standard deviation of 1.3 percent noted.\n\nThis patch:\n\nSimply move update_and_free_page() so that it can be reused later in this\npatch series. The implementation is not changed.\n\nSigned-off-by: Adam Litke \u003cagl@us.ibm.com\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Dave McCracken \u003cdave.mccracken@oracle.com\u003e\nAcked-by: William Irwin \u003cbill.irwin@oracle.com\u003e\nCc: David Gibson \u003cdavid@gibson.dropbear.id.au\u003e\nCc: Ken Chen \u003ckenchen@google.com\u003e\nCc: Badari Pulavarty \u003cpbadari@us.ibm.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "98f3cfc1dc7a53b629d43b7844a9b3f786213048", "tree": "de98ff0e3c25b01863c31abc0f3054de978b9ef8", "parents": [ "48e94196a533dbee17c252bf80d0310fb8c8c2eb" ], "author": { "name": "Yasunori Goto", "email": "y-goto@jp.fujitsu.com", "time": "Tue Oct 16 01:26:14 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:02 2007 -0700" }, "message": "memory hotplug: Hot-add with sparsemem-vmemmap\n\nThis patch is to avoid panic when memory hot-add is executed with\nsparsemem-vmemmap. Current vmemmap-sparsemem code doesn\u0027t support memory\nhot-add. Vmemmap must be populated when hot-add. This is for\n2.6.23-rc2-mm2.\n\nTodo: # Even if this patch is applied, the message \"[xxxx-xxxx] potential\n offnode page_structs\" is displayed. To allocate memmap on its node,\n memmap (and pgdat) must be initialized itself like chicken and\n egg relationship.\n\n # vmemmap_unpopulate will be necessary for followings.\n - For cancel hot-add due to error.\n - For unplug.\n\nSigned-off-by: Yasunori Goto \u003cy-goto@jp.fujitsu.com\u003e\nCc: Andy Whitcroft \u003capw@shadowen.org\u003e\nCc: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "48e94196a533dbee17c252bf80d0310fb8c8c2eb", "tree": "9b29060b0e2600dc7da645fd436dc94c25d1fb36", "parents": [ "de33b821f34dc9e2fabf81a8ba464f3ef039bd00" ], "author": { "name": "KAMEZAWA Hiroyuki", "email": "kamezawa.hiroyu@jp.fujitsu.com", "time": "Tue Oct 16 01:26:14 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:02 2007 -0700" }, "message": "fix memory hot remove not configured case.\n\nNow, arch dependent code around CONFIG_MEMORY_HOTREMOVE is a mess.\nThis patch cleans up them. This is against 2.6.23-rc6-mm1.\n\n - fix compile failure on ia64/ CONFIG_MEMORY_HOTPLUG \u0026\u0026 !CONFIG_MEMORY_HOTREMOVE case.\n - For !CONFIG_MEMORY_HOTREMOVE, add generic no-op remove_memory(),\n which returns -EINVAL.\n - removed remove_pages() only used in powerpc.\n - removed no-op remove_memory() in i386, sh, sparc64, x86_64.\n\n - only powerpc returns -ENOSYS at memory hot remove(no-op). changes it\n to return -EINVAL.\n\nNote:\nCurrently, only ia64 supports CONFIG_MEMORY_HOTREMOVE. I welcome other\narchs if there are requirements and testers.\n\nSigned-off-by: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "0c0e6195896535481173df98935ad8db174f4d45", "tree": "2b35d3b81ba54b5d38e691d2a2019f4bcdfd1dce", "parents": [ "a5d76b54a3f3a40385d7f76069a2feac9f1bad63" ], "author": { "name": "KAMEZAWA Hiroyuki", "email": "kamezawa.hiroyu@jp.fujitsu.com", "time": "Tue Oct 16 01:26:12 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:02 2007 -0700" }, "message": "memory unplug: page offline\n\nLogic.\n - set all pages in [start,end) as isolated migration-type.\n by this, all free pages in the range will be not-for-use.\n - Migrate all LRU pages in the range.\n - Test all pages in the range\u0027s refcnt is zero or not.\n\nTodo:\n - allocate migration destination page from better area.\n - confirm page_count(page)\u003d\u003d 0 \u0026\u0026 PageReserved(page) page is safe to be freed..\n (I don\u0027t like this kind of page but..\n - Find out pages which cannot be migrated.\n - more running tests.\n - Use reclaim for unplugging other memory type area.\n\nSigned-off-by: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: Yasunori Goto \u003cy-goto@jp.fujitsu.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "a5d76b54a3f3a40385d7f76069a2feac9f1bad63", "tree": "f58c432a4224b3be032bd4a4afa79dfa55d198a6", "parents": [ "75884fb1c6388f3713ddcca662f3647b3129aaeb" ], "author": { "name": "KAMEZAWA Hiroyuki", "email": "kamezawa.hiroyu@jp.fujitsu.com", "time": "Tue Oct 16 01:26:11 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:02 2007 -0700" }, "message": "memory unplug: page isolation\n\nImplement generic chunk-of-pages isolation method by using page grouping ops.\n\nThis patch add MIGRATE_ISOLATE to MIGRATE_TYPES. By this\n - MIGRATE_TYPES increases.\n - bitmap for migratetype is enlarged.\n\npages of MIGRATE_ISOLATE migratetype will not be allocated even if it is free.\nBy this, you can isolated *freed* pages from users. How-to-free pages is not\na purpose of this patch. You may use reclaim and migrate codes to free pages.\n\nIf start_isolate_page_range(start,end) is called,\n - migratetype of the range turns to be MIGRATE_ISOLATE if\n its type is MIGRATE_MOVABLE. (*) this check can be updated if other\n memory reclaiming works make progress.\n - MIGRATE_ISOLATE is not on migratetype fallback list.\n - All free pages and will-be-freed pages are isolated.\nTo check all pages in the range are isolated or not, use test_pages_isolated(),\nTo cancel isolation, use undo_isolate_page_range().\n\nChanges V6 -\u003e V7\n - removed unnecessary #ifdef\n\nThere are HOLES_IN_ZONE handling codes...I\u0027m glad if we can remove them..\n\nSigned-off-by: Yasunori Goto \u003cy-goto@jp.fujitsu.com\u003e\nSigned-off-by: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "75884fb1c6388f3713ddcca662f3647b3129aaeb", "tree": "7debdd89fd94d099de3d3763b47af00ef6359d9d", "parents": [ "48f13bf3e742fca8aab87f6c39451d03bf5952d4" ], "author": { "name": "KAMEZAWA Hiroyuki", "email": "kamezawa.hiroyu@jp.fujitsu.com", "time": "Tue Oct 16 01:26:10 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "memory unplug: memory hotplug cleanup\n\nA clean up patch for \"scanning memory resource [start, end)\" operation.\n\nNow, find_next_system_ram() function is used in memory hotplug, but this\ninterface is not easy to use and codes are complicated.\n\nThis patch adds walk_memory_resouce(start,len,arg,func) function.\nThe function \u0027func\u0027 is called per valid memory resouce range in [start,pfn).\n\n[pbadari@us.ibm.com: Error handling in walk_memory_resource()]\nSigned-off-by: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: Badari Pulavarty \u003cpbadari@us.ibm.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "48f13bf3e742fca8aab87f6c39451d03bf5952d4", "tree": "668160019ab157500a90655cf44f798ed3c77893", "parents": [ "ea3061d227816d00717446ac12b853d7ae04b4fe" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:26:10 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "Breakout page_order() to internal.h to avoid special knowledge of the buddy allocator\n\nThe statistics patch later needs to know what order a free page is on the free\nlists. Rather than having special knowledge of page_private() when\nPageBuddy() is set, this patch places out page_order() in internal.h and adds\na VM_BUG_ON to catch using it on non-PageBuddy pages.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "ea3061d227816d00717446ac12b853d7ae04b4fe", "tree": "fa63e359433ba09481d0c758cdf54c877b29e5f0", "parents": [ "42a9fdbb12ac6c027b4b91ab9b5a60aa3a834489" ], "author": { "name": "Andrew Morton", "email": "akpm@linux-foundation.org", "time": "Tue Oct 16 01:26:09 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "slub: list_locations() can use GFP_TEMPORARY\n\nIt\u0027s a short-lived allocation.\n\nCc: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "42a9fdbb12ac6c027b4b91ab9b5a60aa3a834489", "tree": "d319573849af86fa405b93e5b93add0d34305db7", "parents": [ "4c93c355d5d563f300df7e61ef753d7a064411e9" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:26:09 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "SLUB: Optimize cacheline use for zeroing\n\nWe touch a cacheline in the kmem_cache structure for zeroing to get the\nsize. However, the hot paths in slab_alloc and slab_free do not reference\nany other fields in kmem_cache, so we may have to just bring in the\ncacheline for this one access.\n\nAdd a new field to kmem_cache_cpu that contains the object size. That\ncacheline must already be used in the hotpaths. So we save one cacheline\non every slab_alloc if we zero.\n\nWe need to update the kmem_cache_cpu object size if an aliasing operation\nchanges the objsize of an non debug slab.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "4c93c355d5d563f300df7e61ef753d7a064411e9", "tree": "24bcdbed58a51c69640da9c8e220dd5ce0c054a7", "parents": [ "ee3c72a14bfecdf783738032ff3c73ef6412f5b3" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:26:08 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "SLUB: Place kmem_cache_cpu structures in a NUMA aware way\n\nThe kmem_cache_cpu structures introduced are currently an array placed in the\nkmem_cache struct. Meaning the kmem_cache_cpu structures are overwhelmingly\non the wrong node for systems with a higher amount of nodes. These are\nperformance critical structures since the per node information has\nto be touched for every alloc and free in a slab.\n\nIn order to place the kmem_cache_cpu structure optimally we put an array\nof pointers to kmem_cache_cpu structs in kmem_cache (similar to SLAB).\n\nHowever, the kmem_cache_cpu structures can now be allocated in a more\nintelligent way.\n\nWe would like to put per cpu structures for the same cpu but different\nslab caches in cachelines together to save space and decrease the cache\nfootprint. However, the slab allocators itself control only allocations\nper node. We set up a simple per cpu array for every processor with\n100 per cpu structures which is usually enough to get them all set up right.\nIf we run out then we fall back to kmalloc_node. This also solves the\nbootstrap problem since we do not have to use slab allocator functions\nearly in boot to get memory for the small per cpu structures.\n\nPro:\n\t- NUMA aware placement improves memory performance\n\t- All global structures in struct kmem_cache become readonly\n\t- Dense packing of per cpu structures reduces cacheline\n\t footprint in SMP and NUMA.\n\t- Potential avoidance of exclusive cacheline fetches\n\t on the free and alloc hotpath since multiple kmem_cache_cpu\n\t structures are in one cacheline. This is particularly important\n\t for the kmalloc array.\n\nCons:\n\t- Additional reference to one read only cacheline (per cpu\n\t array of pointers to kmem_cache_cpu) in both slab_alloc()\n\t and slab_free().\n\n[akinobu.mita@gmail.com: fix cpu hotplug offline/online path]\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: \"Pekka Enberg\" \u003cpenberg@cs.helsinki.fi\u003e\nCc: Akinobu Mita \u003cakinobu.mita@gmail.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "ee3c72a14bfecdf783738032ff3c73ef6412f5b3", "tree": "876c2c5d33058be8502504330726bb16b876ba52", "parents": [ "b3fba8da653999c67d7517050f196e92da6f8d3b" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:26:07 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "SLUB: Avoid touching page struct when freeing to per cpu slab\n\nSet c-\u003enode to -1 if we allocate from a debug slab instead for SlabDebug\nwhich requires access the page struct cacheline.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nTested-by: Alexey Dobriyan \u003cadobriyan@sw.ru\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "b3fba8da653999c67d7517050f196e92da6f8d3b", "tree": "12ee00ca91ceafe4fa9bb75d62debaf1feebb7b9", "parents": [ "8e65d24c7caf2a4c69b3ae0ce170bf3082ba359f" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:26:06 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "SLUB: Move page-\u003eoffset to kmem_cache_cpu-\u003eoffset\n\nWe need the offset from the page struct during slab_alloc and slab_free. In\nboth cases we also reference the cacheline of the kmem_cache_cpu structure.\nWe can therefore move the offset field into the kmem_cache_cpu structure\nfreeing up 16 bits in the page struct.\n\nMoving the offset allows an allocation from slab_alloc() without touching the\npage struct in the hot path.\n\nThe only thing left in slab_free() that touches the page struct cacheline for\nper cpu freeing is the checking of SlabDebug(page). The next patch deals with\nthat.\n\nUse the available 16 bits to broaden page-\u003einuse. More than 64k objects per\nslab become possible and we can get rid of the checks for that limitation.\n\nNo need anymore to shrink the order of slabs if we boot with 2M sized slabs\n(slub_min_order\u003d9).\n\nNo need anymore to switch off the offset calculation for very large slabs\nsince the field in the kmem_cache_cpu structure is 32 bits and so the offset\nfield can now handle slab sizes of up to 8GB.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "8e65d24c7caf2a4c69b3ae0ce170bf3082ba359f", "tree": "4f690448c1363bf02f74abd9293126c3e3a9e4c9", "parents": [ "dfb4f09609827301740ef0a11b37530d190f1681" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:26:06 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "SLUB: Do not use page-\u003emapping\n\nAfter moving the lockless_freelist to kmem_cache_cpu we no longer need\npage-\u003elockless_freelist. Restructure the use of the struct page fields in\nsuch a way that we never touch the mapping field.\n\nThis is turn allows us to remove the special casing of SLUB when determining\nthe mapping of a page (needed for corner cases of virtual caches machines that\nneed to flush caches of processors mapping a page).\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "dfb4f09609827301740ef0a11b37530d190f1681", "tree": "eb4d13d8699cf01abada9f45e1670cc601fb4b00", "parents": [ "484f51f820199ab3e0ef15d08f1b6be20f53bf39" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:26:05 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "SLUB: Avoid page struct cacheline bouncing due to remote frees to cpu slab\n\nA remote free may access the same page struct that also contains the lockless\nfreelist for the cpu slab. If objects have a short lifetime and are freed by\na different processor then remote frees back to the slab from which we are\ncurrently allocating are frequent. The cacheline with the page struct needs\nto be repeately acquired in exclusive mode by both the allocating thread and\nthe freeing thread. If this is frequent enough then performance will suffer\nbecause of cacheline bouncing.\n\nThis patchset puts the lockless_freelist pointer in its own cacheline. In\norder to make that happen we introduce a per cpu structure called\nkmem_cache_cpu.\n\nInstead of keeping an array of pointers to page structs we now keep an array\nto a per cpu structure that--among other things--contains the pointer to the\nlockless freelist. The freeing thread can then keep possession of exclusive\naccess to the page struct cacheline while the allocating thread keeps its\nexclusive access to the cacheline containing the per cpu structure.\n\nThis works as long as the allocating cpu is able to service its request\nfrom the lockless freelist. If the lockless freelist runs empty then the\nallocating thread needs to acquire exclusive access to the cacheline with\nthe page struct lock the slab.\n\nThe allocating thread will then check if new objects were freed to the per\ncpu slab. If so it will keep the slab as the cpu slab and continue with the\nrecently remote freed objects. So the allocating thread can take a series\nof just freed remote pages and dish them out again. Ideally allocations\ncould be just recycling objects in the same slab this way which will lead\nto an ideal allocation / remote free pattern.\n\nThe number of objects that can be handled in this way is limited by the\ncapacity of one slab. Increasing slab size via slub_min_objects/\nslub_max_order may increase the number of objects and therefore performance.\n\nIf the allocating thread runs out of objects and finds that no objects were\nput back by the remote processor then it will retrieve a new slab (from the\npartial lists or from the page allocator) and start with a whole\nnew set of objects while the remote thread may still be freeing objects to\nthe old cpu slab. This may then repeat until the new slab is also exhausted.\nIf remote freeing has freed objects in the earlier slab then that earlier\nslab will now be on the partial freelist and the allocating thread will\npick that slab next for allocation. So the loop is extended. However,\nboth threads need to take the list_lock to make the swizzling via\nthe partial list happen.\n\nIt is likely that this kind of scheme will keep the objects being passed\naround to a small set that can be kept in the cpu caches leading to increased\nperformance.\n\nMore code cleanups become possible:\n\n- Instead of passing a cpu we can now pass a kmem_cache_cpu structure around.\n Allows reducing the number of parameters to various functions.\n- Can define a new node_match() function for NUMA to encapsulate locality\n checks.\n\nEffect on allocations:\n\nCachelines touched before this patch:\n\n\tWrite:\tpage cache struct and first cacheline of object\n\nCachelines touched after this patch:\n\n\tWrite:\tkmem_cache_cpu cacheline and first cacheline of object\n\tRead: page cache struct (but see later patch that avoids touching\n\t\tthat cacheline)\n\nThe handling when the lockless alloc list runs empty gets to be a bit more\ncomplicated since another cacheline has now to be written to. But that is\nhalfway out of the hot path.\n\nEffect on freeing:\n\nCachelines touched before this patch:\n\n\tWrite: page_struct and first cacheline of object\n\nCachelines touched after this patch depending on how we free:\n\n Write(to cpu_slab):\tkmem_cache_cpu struct and first cacheline of object\n Write(to other):\tpage struct and first cacheline of object\n\n Read(to cpu_slab):\tpage struct to id slab etc. (but see later patch that\n \t\t\tavoids touching the page struct on free)\n Read(to other):\tcpu local kmem_cache_cpu struct to verify its not\n \t\t\tthe cpu slab.\n\nSummary:\n\nPro:\n\t- Distinct cachelines so that concurrent remote frees and local\n\t allocs on a cpuslab can occur without cacheline bouncing.\n\t- Avoids potential bouncing cachelines because of neighboring\n\t per cpu pointer updates in kmem_cache\u0027s cpu_slab structure since\n\t it now grows to a cacheline (Therefore remove the comment\n\t that talks about that concern).\n\nCons:\n\t- Freeing objects now requires the reading of one additional\n\t cacheline. That can be mitigated for some cases by the following\n\t patches but its not possible to completely eliminate these\n\t references.\n\n\t- Memory usage grows slightly.\n\n\tThe size of each per cpu object is blown up from one word\n\t(pointing to the page_struct) to one cacheline with various data.\n\tSo this is NR_CPUS*NR_SLABS*L1_BYTES more memory use. Lets say\n\tNR_SLABS is 100 and a cache line size of 128 then we have just\n\tincreased SLAB metadata requirements by 12.8k per cpu.\n\t(Another later patch reduces these requirements)\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "484f51f820199ab3e0ef15d08f1b6be20f53bf39", "tree": "eea15a3cb546463488595e7a5fae8628a5dd7877", "parents": [ "467c996c1e1910633fa8e7adc9b052aa3ed5f97c" ], "author": { "name": "Adrian Bunk", "email": "bunk@stusta.de", "time": "Tue Oct 16 01:26:03 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:01 2007 -0700" }, "message": "mm/page_alloc.c: make code static\n\nThis patch makes needlessly global code static.\n\nSigned-off-by: Adrian Bunk \u003cbunk@stusta.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "467c996c1e1910633fa8e7adc9b052aa3ed5f97c", "tree": "09e0e70160386be1bdaa12801afddf287e12c8a1", "parents": [ "d9c2340052278d8eb2ffb16b0484f8f794def4de" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:26:02 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Print out statistics in relation to fragmentation avoidance to /proc/pagetypeinfo\n\nThis patch provides fragmentation avoidance statistics via /proc/pagetypeinfo.\n The information is collected only on request so there is no runtime overhead.\n The statistics are in three parts:\n\nThe first part prints information on the size of blocks that pages are\nbeing grouped on and looks like\n\nPage block order: 10\nPages per block: 1024\n\nThe second part is a more detailed version of /proc/buddyinfo and looks like\n\nFree pages count per migrate type at order 0 1 2 3 4 5 6 7 8 9 10\nNode 0, zone DMA, type Unmovable 0 0 0 0 0 0 0 0 0 0 0\nNode 0, zone DMA, type Reclaimable 1 0 0 0 0 0 0 0 0 0 0\nNode 0, zone DMA, type Movable 0 0 0 0 0 0 0 0 0 0 0\nNode 0, zone DMA, type Reserve 0 4 4 0 0 0 0 1 0 1 0\nNode 0, zone Normal, type Unmovable 111 8 4 4 2 3 1 0 0 0 0\nNode 0, zone Normal, type Reclaimable 293 89 8 0 0 0 0 0 0 0 0\nNode 0, zone Normal, type Movable 1 6 13 9 7 6 3 0 0 0 0\nNode 0, zone Normal, type Reserve 0 0 0 0 0 0 0 0 0 0 4\n\nThe third part looks like\n\nNumber of blocks type Unmovable Reclaimable Movable Reserve\nNode 0, zone DMA 0 1 2 1\nNode 0, zone Normal 3 17 94 4\n\nTo walk the zones within a node with interrupts disabled, walk_zones_in_node()\nis introduced and shared between /proc/buddyinfo, /proc/zoneinfo and\n/proc/pagetypeinfo to reduce code duplication. It seems specific to what\nvmstat.c requires but could be broken out as a general utility function in\nmmzone.c if there were other other potential users.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "d9c2340052278d8eb2ffb16b0484f8f794def4de", "tree": "aec7e4e11473a4fcdfd389c718544780a042c6df", "parents": [ "d100313fd615cc30374ff92e0b3facb053838330" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:26:01 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Do not depend on MAX_ORDER when grouping pages by mobility\n\nCurrently mobility grouping works at the MAX_ORDER_NR_PAGES level. This makes\nsense for the majority of users where this is also the huge page size.\nHowever, on platforms like ia64 where the huge page size is runtime\nconfigurable it is desirable to group at a lower order. On x86_64 and\noccasionally on x86, the hugepage size may not always be MAX_ORDER_NR_PAGES.\n\nThis patch groups pages together based on the value of HUGETLB_PAGE_ORDER. It\nuses a compile-time constant if possible and a variable where the huge page\nsize is runtime configurable.\n\nIt is assumed that grouping should be done at the lowest sensible order and\nthat the user would not want to override this. If this is not true,\npage_block order could be forced to a variable initialised via a boot-time\nkernel parameter.\n\nOne potential issue with this patch is that IA64 now parses hugepagesz with\nearly_param() instead of __setup(). __setup() is called after the memory\nallocator has been initialised and the pageblock bitmaps already setup. In\ntests on one IA64 there did not seem to be any problem with using\nearly_param() and in fact may be more correct as it guarantees the parameter\nis handled before the parsing of hugepages\u003d.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "d100313fd615cc30374ff92e0b3facb053838330", "tree": "f0bcd5e3b07bee40a65182c63b54baceca366849", "parents": [ "64c5e135bf5a2a7f0ededb3435a31adbe0202f0c" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:26:00 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Fix calculation in move_freepages_block for counting pages\n\nmove_freepages_block() returns the number of blocks moved. This value is used\nto determine if a block of pages should be stolen for the exclusive use of a\nmigrate type or not. However, the value returned is being used correctly.\nThis patch fixes the calculation to return the number of base pages that have\nbeen moved.\n\nThis should be considered a fix to the patch\nmove-free-pages-between-lists-on-steal.patch\n\nCredit to Andy Whitcroft for spotting the problem.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "64c5e135bf5a2a7f0ededb3435a31adbe0202f0c", "tree": "cb4ff93cbcc3c27176723419a313d7c53545d36b", "parents": [ "ac0e5b7a6b93fb291b01fe1e951e3c16bcdd3503" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:59 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "don\u0027t group high order atomic allocations\n\nGrouping high-order atomic allocations together was intended to allow\nbursty users of atomic allocations to work such as e1000 in situations\nwhere their preallocated buffers were depleted. This did not work in at\nleast one case with a wireless network adapter needing order-1 allocations\nfrequently. To resolve that, the free pages used for min_free_kbytes were\nmoved to separate contiguous blocks with the patch\nbias-the-location-of-pages-freed-for-min_free_kbytes-in-the-same-max_order_nr_pages-blocks.\n\nIt is felt that keeping the free pages in the same contiguous blocks should\nbe sufficient for bursty short-lived high-order atomic allocations to\nsucceed, maybe even with the e1000. Even if there is a failure, increasing\nthe value of min_free_kbytes will free pages as contiguous bloks in\ncontrast to the standard buddy allocator which makes no attempt to keep the\nminimum number of free pages contiguous.\n\nThis patch backs out grouping high order atomic allocations together to\ndetermine if it is really needed or not. If a new report comes in about\nhigh-order atomic allocations failing, the feature can be reintroduced to\ndetermine if it fixes the problem or not. As a side-effect, this patch\nreduces by 1 the number of bits required to track the mobility type of\npages within a MAX_ORDER_NR_PAGES block.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "ac0e5b7a6b93fb291b01fe1e951e3c16bcdd3503", "tree": "732f67c8de6e0d2e001b60c17af9599468b80163", "parents": [ "56fd56b868f19385c50af8941a4c78df433b2d32" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:58 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "remove PAGE_GROUP_BY_MOBILITY\n\nGrouping pages by mobility can be disabled at compile-time. This was\nconsidered undesirable by a number of people. However, in the current stack of\npatches, it is not a simple case of just dropping the configurable patch as it\nwould cause merge conflicts. This patch backs out the configuration option.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "56fd56b868f19385c50af8941a4c78df433b2d32", "tree": "5ea8362e6e141e2d1124d4640811c76489567bc5", "parents": [ "5c0e3066474b57c56ff0d88ca31d95bd14232fee" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:58 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Bias the location of pages freed for min_free_kbytes in the same MAX_ORDER_NR_PAGES blocks\n\nThe standard buddy allocator always favours the smallest block of pages.\nThe effect of this is that the pages free to satisfy min_free_kbytes tends\nto be preserved since boot time at the same location of memory ffor a very\nlong time and as a contiguous block. When an administrator sets the\nreserve at 16384 at boot time, it tends to be the same MAX_ORDER blocks\nthat remain free. This allows the occasional high atomic allocation to\nsucceed up until the point the blocks are split. In practice, it is\ndifficult to split these blocks but when they do split, the benefit of\nhaving min_free_kbytes for contiguous blocks disappears. Additionally,\nincreasing min_free_kbytes once the system has been running for some time\nhas no guarantee of creating contiguous blocks.\n\nOn the other hand, CONFIG_PAGE_GROUP_BY_MOBILITY favours splitting large\nblocks when there are no free pages of the appropriate type available. A\nside-effect of this is that all blocks in memory tends to be used up and\nthe contiguous free blocks from boot time are not preserved like in the\nvanilla allocator. This can cause a problem if a new caller is unwilling\nto reclaim or does not reclaim for long enough.\n\nA failure scenario was found for a wireless network device allocating\norder-1 atomic allocations but the allocations were not intense or frequent\nenough for a whole block of pages to be preserved for MIGRATE_HIGHALLOC.\nThis was reproduced on a desktop by booting with mem\u003d256mb, forcing the\ndriver to allocate at order-1, running a bittorrent client (downloading a\ndebian ISO) and building a kernel with -j2.\n\nThis patch addresses the problem on the desktop machine booted with\nmem\u003d256mb. It works by setting aside a reserve of MAX_ORDER_NR_PAGES\nblocks, the number of which depends on the value of min_free_kbytes. These\nblocks are only fallen back to when there is no other free pages. Then the\nsmallest possible page is used just like the normal buddy allocator instead\nof the largest possible page to preserve contiguous pages The pages in free\nlists in the reserve blocks are never taken for another migrate type. The\nresults is that even if min_free_kbytes is set to a low value, contiguous\nblocks will be preserved in the MIGRATE_RESERVE blocks.\n\nThis works better than the vanilla allocator because if min_free_kbytes is\nincreased, a new reserve block will be chosen based on the location of\nreclaimable pages and the block will free up as contiguous pages. In the\nvanilla allocator, no effort is made to target a block of pages to free as\ncontiguous pages and min_free_kbytes pages are scattered randomly.\n\nThis effect has been observed on the test machine. min_free_kbytes was set\ninitially low but it was kept as a contiguous free block within\nMIGRATE_RESERVE. min_free_kbytes was then set to a higher value and over a\nperiod of time, the free blocks were within the reserve and coalescing.\nHow long it takes to free up depends on how quickly LRU is rotating.\nAmusingly, this means that more activity will free the blocks faster.\n\nThis mechanism potentially replaces MIGRATE_HIGHALLOC as it may be more\neffective than grouping contiguous free pages together. It all depends on\nwhether the number of active atomic high allocations exceeds\nmin_free_kbytes or not. If the number of active allocations exceeds\nmin_free_kbytes, it\u0027s worth it but maybe in that situation, min_free_kbytes\nshould be set higher. Once there are no more reports of allocation\nfailures, a patch will be submitted that backs out MIGRATE_HIGHALLOC and\nsee if the reports stay missing.\n\nCredit to Mariusz Kozlowski for discovering the problem, describing the\nfailure scenario and testing patches and scenarios.\n\n[akpm@linux-foundation.org: cleanups]\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "5c0e3066474b57c56ff0d88ca31d95bd14232fee", "tree": "90c963c62891db4a9039e84e615c01408b09c845", "parents": [ "46dafbca2bba811665b01d8cedf911204820623c" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:56 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Fix corruption of memmap on IA64 SPARSEMEM when mem_section is not a power of 2\n\nThere are problems in the use of SPARSEMEM and pageblock flags that causes\nproblems on ia64.\n\nThe first part of the problem is that units are incorrect in\nSECTION_BLOCKFLAGS_BITS computation. This results in a map_section\u0027s\nsection_mem_map being treated as part of a bitmap which isn\u0027t good. This\nwas evident with an invalid virtual address when mem_init attempted to free\nbootmem pages while relinquishing control from the bootmem allocator.\n\nThe second part of the problem occurs because the pageblock flags bitmap is\nbe located with the mem_section. The SECTIONS_PER_ROOT computation using\nsizeof (mem_section) may not be a power of 2 depending on the size of the\nbitmap. This renders masks and other such things not power of 2 base.\nThis issue was seen with SPARSEMEM_EXTREME on ia64. This patch moves the\nbitmap outside of mem_section and uses a pointer instead in the\nmem_section. The bitmaps are allocated when the section is being\ninitialised.\n\nNote that sparse_early_usemap_alloc() does not use alloc_remap() like\nsparse_early_mem_map_alloc(). The allocation required for the bitmap on\nx86, the only architecture that uses alloc_remap is typically smaller than\na cache line. alloc_remap() pads out allocations to the cache size which\nwould be a needless waste.\n\nCredit to Bob Picco for identifying the original problem and effecting a\nfix for the SECTION_BLOCKFLAGS_BITS calculation. Credit to Andy Whitcroft\nfor devising the best way of allocating the bitmaps only when required for\nthe section.\n\n[wli@holomorphy.com: warning fix]\nSigned-off-by: Bob Picco \u003cbob.picco@hp.com\u003e\nSigned-off-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nCc: \"Luck, Tony\" \u003ctony.luck@intel.com\u003e\nSigned-off-by: William Irwin \u003cbill.irwin@oracle.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "46dafbca2bba811665b01d8cedf911204820623c", "tree": "c0dbc78b3da9749f257fa8756cd3133f39cd4f88", "parents": [ "5adc5be7cd1bcef6bb64f5255d2a33f20a3cf5be" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:55 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Be more agressive about stealing when MIGRATE_RECLAIMABLE allocations fallback\n\nMIGRATE_RECLAIMABLE allocations tend to be very bursty in nature like when\nupdatedb starts. It is likely this will occur in situations where MAX_ORDER\nblocks of pages are not free. This means that updatedb can scatter\nMIGRATE_RECLAIMABLE pages throughout the address space. This patch is more\nagressive about stealing blocks of pages for MIGRATE_RECLAIMABLE.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "5adc5be7cd1bcef6bb64f5255d2a33f20a3cf5be", "tree": "39ed023e8c36cde0c0f9cba50f945e77d3ca26fa", "parents": [ "9ef9acb05a741ec10a5e9122717736de12adced9" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:54 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Bias the placement of kernel pages at lower PFNs\n\nThis patch chooses blocks with lower PFNs when placing kernel allocations.\nThis is particularly important during fallback in low memory situations to\nstop unmovable pages being placed throughout the entire address space.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "9ef9acb05a741ec10a5e9122717736de12adced9", "tree": "6008083999b3c6e115714fcdea637195f2b53df6", "parents": [ "e010487dbe09d63cf916fd1b119d17abd0f48207" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:54 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Do not group pages by mobility type on low memory systems\n\nGrouping pages by mobility can only successfully operate when there are more\nMAX_ORDER_NR_PAGES areas than mobility types. When there are insufficient\nareas, fallbacks cannot be avoided. This has noticeable performance impacts\non machines with small amounts of memory in comparison to MAX_ORDER_NR_PAGES.\nFor example, on IA64 with a configuration including huge pages spans 1GiB with\nMAX_ORDER_NR_PAGES so would need at least 4GiB of RAM before grouping pages by\nmobility would be useful. In comparison, an x86 would need 16MB.\n\nThis patch checks the size of vm_total_pages in build_all_zonelists(). If\nthere are not enough areas, mobility is effectivly disabled by considering\nall allocations as the same type (UNMOVABLE). This is achived via a\n__read_mostly flag.\n\nWith this patch, performance is comparable to disabling grouping pages\nby mobility at compile-time on a test machine with insufficient memory.\nWith this patch, it is reasonable to get rid of grouping pages by mobility\na compile-time option.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nAcked-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "e010487dbe09d63cf916fd1b119d17abd0f48207", "tree": "37c7f36913daf4bc0a68a1d0ba1cc30ee0d4e307", "parents": [ "e12ba74d8ff3e2f73a583500d7095e406df4d093" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:53 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Group high-order atomic allocations\n\nIn rare cases, the kernel needs to allocate a high-order block of pages\nwithout sleeping. For example, this is the case with e1000 cards configured\nto use jumbo frames. Migrating or reclaiming pages in this situation is not\nan option.\n\nThis patch groups these allocations together as much as possible by adding a\nnew MIGRATE_TYPE. The MIGRATE_HIGHATOMIC type are exactly what they sound\nlike. Care is taken that pages of other migrate types do not use the same\nblocks as high-order atomic allocations.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "e12ba74d8ff3e2f73a583500d7095e406df4d093", "tree": "a0d3385b65f0b3e1e00b0bbf11b75e7538a93edb", "parents": [ "c361be55b3128474aa66d31092db330b07539103" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:52 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Group short-lived and reclaimable kernel allocations\n\nThis patch marks a number of allocations that are either short-lived such as\nnetwork buffers or are reclaimable such as inode allocations. When something\nlike updatedb is called, long-lived and unmovable kernel allocations tend to\nbe spread throughout the address space which increases fragmentation.\n\nThis patch groups these allocations together as much as possible by adding a\nnew MIGRATE_TYPE. The MIGRATE_RECLAIMABLE type is for allocations that can be\nreclaimed on demand, but not moved. i.e. they can be migrated by deleting\nthem and re-reading the information from elsewhere.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nCc: Andy Whitcroft \u003capw@shadowen.org\u003e\nCc: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "c361be55b3128474aa66d31092db330b07539103", "tree": "9ce134f4e679144d28f5c32924bdba999a1aae6b", "parents": [ "e2c55dc87f4a398b9c4dcc702dbc23a07fe14e23" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:51 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:43:00 2007 -0700" }, "message": "Move free pages between lists on steal\n\nWhen a fallback occurs, there will be free pages for one allocation type\nstored on the list for another. When a large steal occurs, this patch will\nmove all the free pages within one list to the other.\n\n[y-goto@jp.fujitsu.com: fix BUG_ON check at move_freepages()]\n[apw@shadowen.org: Move to using pfn_valid_within()]\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nCc: Christoph Lameter \u003cclameter@engr.sgi.com\u003e\nSigned-off-by: Yasunori Goto \u003cy-goto@jp.fujitsu.com\u003e\nCc: Bjorn Helgaas \u003cbjorn.helgaas@hp.com\u003e\nSigned-off-by: Andy Whitcroft \u003candyw@uk.ibm.com\u003e\nCc: Bob Picco \u003cbob.picco@hp.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "e2c55dc87f4a398b9c4dcc702dbc23a07fe14e23", "tree": "b3811de298d4c98c4765db4af3838428553b5382", "parents": [ "b92a6edd4b77a8794adb497280beea5df5e59a14" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:50 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "Drain per-cpu lists when high-order allocations fail\n\nPer-cpu pages can accidentally cause fragmentation because they are free, but\npinned pages in an otherwise contiguous block. When this patch is applied,\nthe per-cpu caches are drained after the direct-reclaim is entered if the\nrequested order is greater than 0. It simply reuses the code used by suspend\nand hotplug.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "b92a6edd4b77a8794adb497280beea5df5e59a14", "tree": "396ea5cf2b53fc066e949c443f03747ec868de1e", "parents": [ "535131e6925b4a95f321148ad7293f496e0e58d7" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:50 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "Add a configure option to group pages by mobility\n\nThe grouping mechanism has some memory overhead and a more complex allocation\npath. This patch allows the strategy to be disabled for small memory systems\nor if it is known the workload is suffering because of the strategy. It also\nacts to show where the page groupings strategy interacts with the standard\nbuddy allocator.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Joel Schopp \u003cjschopp@austin.ibm.com\u003e\nCc: Andy Whitcroft \u003capw@shadowen.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "535131e6925b4a95f321148ad7293f496e0e58d7", "tree": "fdd49e29f89eb6db3ba2b5ba7df7b059de95a91f", "parents": [ "b2a0ac8875a0a3b9f0739b60526f8c5977d2200f" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:49 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "Choose pages from the per-cpu list based on migration type\n\nThe freelists for each migrate type can slowly become polluted due to the\nper-cpu list. Consider what happens when the following happens\n\n1. A 2^(MAX_ORDER-1) list is reserved for __GFP_MOVABLE pages\n2. An order-0 page is allocated from the newly reserved block\n3. The page is freed and placed on the per-cpu list\n4. alloc_page() is called with GFP_KERNEL as the gfp_mask\n5. The per-cpu list is used to satisfy the allocation\n\nThis results in a kernel page is in the middle of a migratable region. This\npatch prevents this leak occuring by storing the MIGRATE_ type of the page in\npage-\u003eprivate. On allocate, a page will only be returned of the desired type,\nelse more pages will be allocated. This may temporarily allow a per-cpu list\nto go over the pcp-\u003ehigh limit but it\u0027ll be corrected on the next free. Care\nis taken to preserve the hotness of pages recently freed.\n\nThe additional code is not measurably slower for the workloads we\u0027ve tested.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "b2a0ac8875a0a3b9f0739b60526f8c5977d2200f", "tree": "31826716b3209751a5468b840ff14190b4a5a8a2", "parents": [ "835c134ec4dd755e5c4470af566db226d1e96742" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:48 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "Split the free lists for movable and unmovable allocations\n\nThis patch adds the core of the fragmentation reduction strategy. It works by\ngrouping pages together based on their ability to migrate or be reclaimed.\nBasically, it works by breaking the list in zone-\u003efree_area list into\nMIGRATE_TYPES number of lists.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "835c134ec4dd755e5c4470af566db226d1e96742", "tree": "7bc659978b4fba5089fc820185a8b6f0cc010b08", "parents": [ "954ffcb35f5aca428661d29b96c4eee82b3c19cd" ], "author": { "name": "Mel Gorman", "email": "mel@csn.ul.ie", "time": "Tue Oct 16 01:25:47 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "Add a bitmap that is used to track flags affecting a block of pages\n\nHere is the latest revision of the anti-fragmentation patches. Of particular\nnote in this version is special treatment of high-order atomic allocations.\nCare is taken to group them together and avoid grouping pages of other types\nnear them. Artifical tests imply that it works. I\u0027m trying to get the\nhardware together that would allow setting up of a \"real\" test. If anyone\nalready has a setup and test that can trigger the atomic-allocation problem,\nI\u0027d appreciate a test of these patches and a report. The second major change\nis that these patches will apply cleanly with patches that implement\nanti-fragmentation through zones.\n\nkernbench shows effectively no performance difference varying between -0.2%\nand +2% on a variety of test machines. Success rates for huge page allocation\nare dramatically increased. For example, on a ppc64 machine, the vanilla\nkernel was only able to allocate 1% of memory as a hugepage and this was due\nto a single hugepage reserved as min_free_kbytes. With these patches applied,\n17% was allocatable as superpages. With reclaim-related fixes from Andy\nWhitcroft, it was 40% and further reclaim-related improvements should increase\nthis further.\n\nChangelog Since V28\no Group high-order atomic allocations together\no It is no longer required to set min_free_kbytes to 10% of memory. A value\n of 16384 in most cases will be sufficient\no Now applied with zone-based anti-fragmentation\no Fix incorrect VM_BUG_ON within buffered_rmqueue()\no Reorder the stack so later patches do not back out work from earlier patches\no Fix bug were journal pages were being treated as movable\no Bias placement of non-movable pages to lower PFNs\no More agressive clustering of reclaimable pages in reactions to workloads\n like updatedb that flood the size of inode caches\n\nChangelog Since V27\n\no Renamed anti-fragmentation to Page Clustering. Anti-fragmentation was giving\n the mistaken impression that it was the 100% solution for high order\n allocations. Instead, it greatly increases the chances high-order\n allocations will succeed and lays the foundation for defragmentation and\n memory hot-remove to work properly\no Redefine page groupings based on ability to migrate or reclaim instead of\n basing on reclaimability alone\no Get rid of spurious inits\no Per-cpu lists are no longer split up per-type. Instead the per-cpu list is\n searched for a page of the appropriate type\no Added more explanation commentary\no Fix up bug in pageblock code where bitmap was used before being initalised\n\nChangelog Since V26\no Fix double init of lists in setup_pageset\n\nChangelog Since V25\no Fix loop order of for_each_rclmtype_order so that order of loop matches args\no gfpflags_to_rclmtype uses gfp_t instead of unsigned long\no Rename get_pageblock_type() to get_page_rclmtype()\no Fix alignment problem in move_freepages()\no Add mechanism for assigning flags to blocks of pages instead of page-\u003eflags\no On fallback, do not examine the preferred list of free pages a second time\n\nThe purpose of these patches is to reduce external fragmentation by grouping\npages of related types together. When pages are migrated (or reclaimed under\nmemory pressure), large contiguous pages will be freed.\n\nThis patch works by categorising allocations by their ability to migrate;\n\nMovable - The pages may be moved with the page migration mechanism. These are\n\tgenerally userspace pages.\n\nReclaimable - These are allocations for some kernel caches that are\n\treclaimable or allocations that are known to be very short-lived.\n\nUnmovable - These are pages that are allocated by the kernel that\n\tare not trivially reclaimed. For example, the memory allocated for a\n\tloaded module would be in this category. By default, allocations are\n\tconsidered to be of this type\n\nHighAtomic - These are high-order allocations belonging to callers that\n\tcannot sleep or perform any IO. In practice, this is restricted to\n\tjumbo frame allocation for network receive. It is assumed that the\n\tallocations are short-lived\n\nInstead of having one MAX_ORDER-sized array of free lists in struct free_area,\nthere is one for each type of reclaimability. Once a 2^MAX_ORDER block of\npages is split for a type of allocation, it is added to the free-lists for\nthat type, in effect reserving it. Hence, over time, pages of the different\ntypes can be clustered together.\n\nWhen the preferred freelists are expired, the largest possible block is taken\nfrom an alternative list. Buddies that are split from that large block are\nplaced on the preferred allocation-type freelists to mitigate fragmentation.\n\nThis implementation gives best-effort for low fragmentation in all zones.\nIdeally, min_free_kbytes needs to be set to a value equal to 4 * (1 \u003c\u003c\n(MAX_ORDER-1)) pages in most cases. This would be 16384 on x86 and x86_64 for\nexample.\n\nOur tests show that about 60-70% of physical memory can be allocated on a\ndesktop after a few days uptime. In benchmarks and stress tests, we are\nfinding that 80% of memory is available as contiguous blocks at the end of the\ntest. To compare, a standard kernel was getting \u003c 1% of memory as large pages\non a desktop and about 8-12% of memory as large pages at the end of stress\ntests.\n\nFollowing this email are 12 patches that implement thie page grouping feature.\n The first patch introduces a mechanism for storing flags related to a whole\nblock of pages. Then allocations are split between movable and all other\nallocations. Following that are patches to deal with per-cpu pages and make\nthe mechanism configurable. The next patch moves free pages between lists\nwhen partially allocated blocks are used for pages of another migrate type.\nThe second last patch groups reclaimable kernel allocations such as inode\ncaches together. The final patch related to groupings keeps high-order atomic\nallocations.\n\nThe last two patches are more concerned with control of fragmentation. The\nsecond last patch biases placement of non-movable allocations towards the\nstart of memory. This is with a view of supporting memory hot-remove of DIMMs\nwith higher PFNs in the future. The biasing could be enforced a lot heavier\nbut it would cost. The last patch agressively clusters reclaimable pages like\ninode caches together.\n\nThe fragmentation reduction strategy needs to track if pages within a block\ncan be moved or reclaimed so that pages are freed to the appropriate list.\nThis patch adds a bitmap for flags affecting a whole a MAX_ORDER block of\npages.\n\nIn non-SPARSEMEM configurations, the bitmap is stored in the struct zone and\nallocated during initialisation. SPARSEMEM statically allocates the bitmap in\na struct mem_section so that bitmaps do not have to be resized during memory\nhotadd. This wastes a small amount of memory per unused section (usually\nsizeof(unsigned long)) but the complexity of dynamically allocating the memory\nis quite high.\n\nAdditional credit to Andy Whitcroft who reviewed up an earlier implementation\nof the mechanism an suggested how to make it a *lot* cleaner.\n\nSigned-off-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nCc: Andy Whitcroft \u003capw@shadowen.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "954ffcb35f5aca428661d29b96c4eee82b3c19cd", "tree": "2dd8aaf26a8ae81b461b6d5d824ae8744690e483", "parents": [ "97ee052461446526e1de7236497e6f1b1ffedf8c" ], "author": { "name": "KAMEZAWA Hiroyuki", "email": "kamezawa.hiroyu@jp.fujitsu.com", "time": "Tue Oct 16 01:25:44 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "flush icache before set_pte() on ia64: flush icache at set_pte\n\nCurrent ia64 kernel flushes icache by lazy_mmu_prot_update() *after*\nset_pte(). This is too late. This patch removes lazy_mmu_prot_update and\nadd modfied set_pte() for flushing if necessary.\n\nThis patch flush icache of a page when\n\tnew pte has exec bit.\n\t\u0026\u0026 new pte has present bit\n\t\u0026\u0026 new pte is user\u0027s page.\n\t\u0026\u0026 (old *ptep is not present\n || new pte\u0027s pfn is not same to old *ptep\u0027s ptn)\n\t\u0026\u0026 new pte\u0027s page has no Pg_arch_1 bit.\n\t Pg_arch_1 is set when a page is cache consistent.\n\nI think this condition checks are much easier to understand than considering\n\"Where sync_icache_dcache() should be inserted ?\".\n\npte_user() for ia64 was removed by http://lkml.org/lkml/2007/6/12/67 as\nclean-up. So, I added it again.\n\nSigned-off-by: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: \"Luck, Tony\" \u003ctony.luck@intel.com\u003e\nCc: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: Hugh Dickins \u003chugh@veritas.com\u003e\nCc: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nAcked-by: David S. Miller \u003cdavem@davemloft.net\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "97ee052461446526e1de7236497e6f1b1ffedf8c", "tree": "7c91270577191ed32de22d8183695c1354d555c8", "parents": [ "4106f83a9f86afc423557d0d92ebf4b3f36728c1" ], "author": { "name": "KAMEZAWA Hiroyuki", "email": "kamezawa.hiroyu@jp.fujitsu.com", "time": "Tue Oct 16 01:25:43 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "flush cache before installing new page at migraton\n\nIn migration, a new page should be cache flushed before set_pte() in some\narchs which have virtually-tagged cache.\n\nSigned-off-by: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: \"Luck, Tony\" \u003ctony.luck@intel.com\u003e\nCc: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: Hugh Dickins \u003chugh@veritas.com\u003e\nCc: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nAcked-by: David S. Miller \u003cdavem@davemloft.net\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "4106f83a9f86afc423557d0d92ebf4b3f36728c1", "tree": "b2da62d411e720b6053a80074c9fb8343ec17ccc", "parents": [ "6cb062296f73e74768cca2f3eaf90deac54de02d" ], "author": { "name": "Andrea Arcangeli", "email": "andrea@suse.de", "time": "Tue Oct 16 01:25:42 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "make swappiness safer to use\n\nSwappiness isn\u0027t a safe sysctl. Setting it to 0 for example can hang a\nsystem. That\u0027s a corner case but even setting it to 10 or lower can waste\nenormous amounts of cpu without making much progress. We\u0027ve customers who\nwants to use swappiness but they can\u0027t because of the current\nimplementation (if you change it so the system stops swapping it really\nstops swapping and nothing works sane anymore if you really had to swap\nsomething to make progress).\n\nThis patch from Kurt Garloff makes swappiness safer to use (no more huge\ncpu usage or hangs with low swappiness values).\n\nI think the prev_priority can also be nuked since it wastes 4 bytes per\nzone (that would be an incremental patch but I wait the nr_scan_[in]active\nto be nuked first for similar reasons). Clearly somebody at some point\nnoticed how broken that thing was and they had to add min(priority,\nprev_priority) to give it some reliability, but they didn\u0027t go the last\nmile to nuke prev_priority too. Calculating distress only in function of\nnot-racy priority is correct and sure more than enough without having to\nadd randomness into the equation.\n\nPatch is tested on older kernels but it compiles and it\u0027s quite simple\nso...\n\nOverall I\u0027m not very satisified by the swappiness tweak, since it doesn\u0027t\nrally do anything with the dirty pagecache that may be inactive. We need\nanother kind of tweak that controls the inactive scan and tunes the\ncan_writepage feature (not yet in mainline despite having submitted it a\nfew times), not only the active one. That new tweak will tell the kernel\nhow hard to scan the inactive list for pure clean pagecache (something the\nmainline kernel isn\u0027t capable of yet). We already have that feature\nworking in all our enterprise kernels with the default reasonable tune, or\nthey can\u0027t even run a readonly backup with tar without triggering huge\nwrite I/O. I think it should be available also in mainline later.\n\nCc: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Kurt Garloff \u003cgarloff@suse.de\u003e\nSigned-off-by: Andrea Arcangeli \u003candrea@suse.de\u003e\nSigned-off-by: Fengguang Wu \u003cwfg@mail.ustc.edu.cn\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6cb062296f73e74768cca2f3eaf90deac54de02d", "tree": "1572139653a6fc97cdffd06f2c1bfa650da2ce03", "parents": [ "58c0a4a7864b2dad6da4090813322fcd29a11c92" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:41 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "Categorize GFP flags\n\nThe function of GFP_LEVEL_MASK seems to be unclear. In order to clear up\nthe mystery we get rid of it and replace GFP_LEVEL_MASK with 3 sets of GFP\nflags:\n\nGFP_RECLAIM_MASK\tFlags used to control page allocator reclaim behavior.\n\nGFP_CONSTRAINT_MASK\tFlags used to limit where allocations can occur.\n\nGFP_SLAB_BUG_MASK\tFlags that the slab allocator BUG()s on.\n\nThese replace the uses of GFP_LEVEL mask in the slab allocators and in\nvmalloc.c.\n\nThe use of the flags not included in these sets may occur as a result of a\nslab allocation standing in for a page allocation when constructing scatter\ngather lists. Extraneous flags are cleared and not passed through to the\npage allocator. __GFP_MOVABLE/RECLAIMABLE, __GFP_COLD and __GFP_COMP will\nnow be ignored if passed to a slab allocator.\n\nChange the allocation of allocator meta data in SLAB and vmalloc to not\npass through flags listed in GFP_CONSTRAINT_MASK. SLAB already removes the\n__GFP_THISNODE flag for such allocations. Generalize that to also cover\nvmalloc. The use of GFP_CONSTRAINT_MASK also includes __GFP_HARDWALL.\n\nThe impact of allocator metadata placement on access latency to the\ncachelines of the object itself is minimal since metadata is only\nreferenced on alloc and free. The attempt is still made to place the meta\ndata optimally but we consistently allow fallback both in SLAB and vmalloc\n(SLUB does not need to allocate metadata like that).\n\nAllocator metadata may serve multiple in kernel users and thus should not\nbe subject to the limitations arising from a single allocation context.\n\n[akpm@linux-foundation.org: fix fallback_alloc()]\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "58c0a4a7864b2dad6da4090813322fcd29a11c92", "tree": "239da4074ced93ee6918fdb03e3839ef721191bb", "parents": [ "37b07e4163f7306aa735a6e250e8d22293e5b8de" ], "author": { "name": "Yasunori Goto", "email": "y-goto@jp.fujitsu.com", "time": "Tue Oct 16 01:25:40 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "Fix panic of cpu online with memory less node\n\nWhen a cpu is onlined on memory-less-node box, kernel panics due to touch\nNULL pointer of pgdat-\u003ekswapd. Current kswapd runs only nodes which have\nmemory. So, calling of set_cpus_allowed() is not necessary for memory-less\nnode.\n\nThis is fix for it.\n\nSigned-off-by: Yasunori Goto \u003cy-goto@jp.fujitsu.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "37b07e4163f7306aa735a6e250e8d22293e5b8de", "tree": "5c9c1935253a39aa840a9923bf1c86620cb6f733", "parents": [ "0e1e7c7a739562a321fda07c7cd2a97a7114f8f8" ], "author": { "name": "Lee Schermerhorn", "email": "Lee.Schermerhorn@hp.com", "time": "Tue Oct 16 01:25:39 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "memoryless nodes: fixup uses of node_online_map in generic code\n\nHere\u0027s a cut at fixing up uses of the online node map in generic code.\n\nmm/shmem.c:shmem_parse_mpol()\n\n\tEnsure nodelist is subset of nodes with memory.\n\tUse node_states[N_HIGH_MEMORY] as default for missing\n\tnodelist for interleave policy.\n\nmm/shmem.c:shmem_fill_super()\n\n\tinitialize policy_nodes to node_states[N_HIGH_MEMORY]\n\nmm/page-writeback.c:highmem_dirtyable_memory()\n\n\tsum over nodes with memory\n\nmm/page_alloc.c:zlc_setup()\n\n\tallowednodes - use nodes with memory.\n\nmm/page_alloc.c:default_zonelist_order()\n\n\taverage over nodes with memory.\n\nmm/page_alloc.c:find_next_best_node()\n\n\tskip nodes w/o memory.\n\tN_HIGH_MEMORY state mask may not be initialized at this time,\n\tunless we want to depend on early_calculate_totalpages() [see\n\tbelow]. Will ZONE_MOVABLE ever be configurable?\n\nmm/page_alloc.c:find_zone_movable_pfns_for_nodes()\n\n\tspread kernelcore over nodes with memory.\n\n\tThis required calling early_calculate_totalpages()\n\tunconditionally, and populating N_HIGH_MEMORY node\n\tstate therein from nodes in the early_node_map[].\n\tIf we can depend on this, we can eliminate the\n\tpopulation of N_HIGH_MEMORY mask from __build_all_zonelists()\n\tand use the N_HIGH_MEMORY mask in find_next_best_node().\n\nmm/mempolicy.c:mpol_check_policy()\n\n\tEnsure nodes specified for policy are subset of\n\tnodes with memory.\n\n[akpm@linux-foundation.org: fix warnings]\nSigned-off-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: Shaohua Li \u003cshaohua.li@intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "523b945855a1427000ffc707c610abe5947ae607", "tree": "2d84b5b6822a2a20bfd79146c08ce06ac8c80b9b", "parents": [ "633c0666b5a5c41c376a5a7e4304d638dc48c1b9" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:37 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "Memoryless nodes: Fix GFP_THISNODE behavior\n\nGFP_THISNODE checks that the zone selected is within the pgdat (node) of the\nfirst zone of a nodelist. That only works if the node has memory. A\nmemoryless node will have its first node on another pgdat (node).\n\nGFP_THISNODE currently will return simply memory on the first pgdat. Thus it\nis returning memory on other nodes. GFP_THISNODE should fail if there is no\nlocal memory on a node.\n\nAdd a new set of zonelists for each node that only contain the nodes that\nbelong to the zones itself so that no fallback is possible.\n\nThen modify gfp_type to pickup the right zone based on the presence of\n__GFP_THISNODE.\n\nDrop the existing GFP_THISNODE checks from the page_allocators hot path.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nAcked-by: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nTested-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "633c0666b5a5c41c376a5a7e4304d638dc48c1b9", "tree": "a8f42f5d4aaa4d25786d5a60e9c92b6d727254fd", "parents": [ "37c0708dbee5825df3bd9ce6ef2199c6c1713970" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:37 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:59 2007 -0700" }, "message": "Memoryless nodes: drop one memoryless node boot warning\n\nget_pfn_range_for_nid() is called multiple times for each node at boot time.\nEach time, it will warn about nodes with no memory, resulting in boot messages\nlike:\n\n Node 0 active with no memory\n Node 0 active with no memory\n Node 0 active with no memory\n Node 0 active with no memory\n Node 0 active with no memory\n Node 0 active with no memory\n On node 0 totalpages: 0\n Node 0 active with no memory\n Node 0 active with no memory\n DMA zone: 0 pages used for memmap\n Node 0 active with no memory\n Node 0 active with no memory\n Normal zone: 0 pages used for memmap\n Node 0 active with no memory\n Node 0 active with no memory\n Movable zone: 0 pages used for memmap\n\nand so on for each memoryless node.\n\nWe already have the \"On node N totalpages: ...\" and other related messages, so\ndrop the \"Node N active with no memory\" warnings.\n\nSigned-off-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nCc: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "37c0708dbee5825df3bd9ce6ef2199c6c1713970", "tree": "747551aa58484e7f872da118b864c8f3ca6e892d", "parents": [ "56bbd65df0e92a4a8eb70c5f2b416ae2b6c5fb31" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:36 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "Memoryless nodes: Add N_CPU node state\n\nWe need the check for a node with cpu in zone reclaim. Zone reclaim will not\nallow remote zone reclaim if a node has a cpu.\n\n[Lee.Schermerhorn@hp.com: Move setup of N_CPU node state mask]\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nTested-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "56bbd65df0e92a4a8eb70c5f2b416ae2b6c5fb31", "tree": "714154b7b16d2e08c60d49b925aa0e789f0f0be0", "parents": [ "4199cfa02b982f4c739e8a6a304d6a40e1935d25" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:35 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "Memoryless nodes: Update memory policy and page migration\n\nOnline nodes now may have no memory. The checks and initialization must\ntherefore be changed to no longer use the online functions.\n\nThis will correctly initialize the interleave on bootup to only target nodes\nwith memory and will make sys_move_pages return an error when a page is to be\nmoved to a memoryless node. Similarly we will get an error if MPOL_BIND and\nMPOL_INTERLEAVE is used on a memoryless node.\n\nThese are somewhat new semantics. So far one could specify memoryless nodes\nand we would maybe do the right thing and just ignore the node (or we\u0027d do\nsomething strange like with MPOL_INTERLEAVE). If we want to allow the\nspecification of memoryless nodes via memory policies then we need to keep\nchecking for online nodes.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nAcked-by: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nTested-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "f64dc58c5412233d4d44b0275eaebdc11bde23b3", "tree": "c35755b1a2a04eefd97d3561640992e5d817ccba", "parents": [ "04231b3002ac53f8a64a7bd142fde3fa4b6808c6" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:33 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "Memoryless nodes: SLUB support\n\nSimply switch all for_each_online_node to for_each_node_state(NORMAL_MEMORY).\nThat way SLUB only operates on nodes with regular memory. Any allocation\nattempt on a memoryless node or a node with just highmem will fall whereupon\nSLUB will fetch memory from a nearby node (depending on how memory policies\nand cpuset describe fallback).\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nTested-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "04231b3002ac53f8a64a7bd142fde3fa4b6808c6", "tree": "7c55f86dbe697621943176cfa2a341dc0e6760ef", "parents": [ "9422ffba4adc82b4b67a3ca6ef51516aa61f8248" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:32 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "Memoryless nodes: Slab support\n\nSlab should not allocate control structures for nodes without memory. This\nmay seem to work right now but its unreliable since not all allocations can\nfall back due to the use of GFP_THISNODE.\n\nSwitching a few for_each_online_node\u0027s to N_NORMAL_MEMORY will allow us to\nonly allocate for nodes that have regular memory.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nAcked-by: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nAcked-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "9422ffba4adc82b4b67a3ca6ef51516aa61f8248", "tree": "e30f87beca455b3e3b1f4c8c8921e9ef049c000f", "parents": [ "ee31af5d649d8aa6ac7948a6d97ae48367ff2d7e" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:31 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "Memoryless nodes: No need for kswapd\n\nA node without memory does not need a kswapd. So use the memory map instead\nof the online map when starting kswapd.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nAcked-by: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nTested-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "ee31af5d649d8aa6ac7948a6d97ae48367ff2d7e", "tree": "8a16d8c15b8b56058b905c2b83a3a5f8112746c4", "parents": [ "6eaf806a223e61dc5f2de4ab591f11beb97a8f3b" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:30 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "Memoryless nodes: OOM: use N_HIGH_MEMORY map instead of constructing one on the fly\n\nconstrained_alloc() builds its own memory map for nodes with memory. We have\nthat available in N_HIGH_MEMORY now. So simplify the code.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nAcked-by: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nAcked-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6eaf806a223e61dc5f2de4ab591f11beb97a8f3b", "tree": "fd63b422c3cb357792d12f9170c354e25b94daa0", "parents": [ "7ea1530ab3fdfa85441061909cc8040e84776fd4" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:30 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "Memoryless nodes: Fix interleave behavior for memoryless nodes\n\nMPOL_INTERLEAVE currently simply loops over all nodes. Allocations on\nmemoryless nodes will be redirected to nodes with memory. This results in an\nimbalance because the neighboring nodes to memoryless nodes will get\nsignificantly more interleave hits that the rest of the nodes on the system.\n\nWe can avoid this imbalance by clearing the nodes in the interleave node set\nthat have no memory. If we use the node map of the memory nodes instead of\nthe online nodes then we have only the nodes we want.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nTested-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "7ea1530ab3fdfa85441061909cc8040e84776fd4", "tree": "f3af0f8ed40a6df90bdbb9396d6163d59798a821", "parents": [ "13808910713a98cc1159291e62cdfec92cc94d05" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:29 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "Memoryless nodes: introduce mask of nodes with memory\n\nIt is necessary to know if nodes have memory since we have recently begun to\nadd support for memoryless nodes. For that purpose we introduce a two new\nnode states: N_HIGH_MEMORY and N_NORMAL_MEMORY.\n\nA node has its bit in N_HIGH_MEMORY set if it has any memory regardless of the\ntype of mmemory. If a node has memory then it has at least one zone defined\nin its pgdat structure that is located in the pgdat itself.\n\nA node has its bit in N_NORMAL_MEMORY set if it has a lower zone than\nZONE_HIGHMEM. This means it is possible to allocate memory that is not\nsubject to kmap.\n\nN_HIGH_MEMORY and N_NORMAL_MEMORY can then be used in various places to insure\nthat we do the right thing when we encounter a memoryless node.\n\n[akpm@linux-foundation.org: build fix]\n[Lee.Schermerhorn@hp.com: update N_HIGH_MEMORY node state for memory hotadd]\n[y-goto@jp.fujitsu.com: Fix memory hotplug + sparsemem build]\nSigned-off-by: Lee Schermerhorn \u003cLee.Schermerhorn@hp.com\u003e\nSigned-off-by: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Yasunori Goto \u003cy-goto@jp.fujitsu.com\u003e\nSigned-off-by: Paul Mundt \u003clethal@linux-sh.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "13808910713a98cc1159291e62cdfec92cc94d05", "tree": "0fd7189dc2a76e1ae165ca5d6e8c6b4e6f1761af", "parents": [ "55144768e100b68447f44c5e5c9deb155ad661bd" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:25:27 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "Memoryless nodes: Generic management of nodemasks for various purposes\n\nWhy do we need to support memoryless nodes?\n\nKAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e wrote:\n\n\u003e For fujitsu, problem is called \"empty\" node.\n\u003e\n\u003e When ACPI\u0027s SRAT table includes \"possible nodes\", ia64 bootstrap(acpi_numa_init)\n\u003e creates nodes, which includes no memory, no cpu.\n\u003e\n\u003e I tried to remove empty-node in past, but that was denied.\n\u003e It was because we can hot-add cpu to the empty node.\n\u003e (node-hotplug triggered by cpu is not implemented now. and it will be ugly.)\n\u003e\n\u003e\n\u003e For HP, (Lee can comment on this later), they have memory-less-node.\n\u003e As far as I hear, HP\u0027s machine can have following configration.\n\u003e\n\u003e (example)\n\u003e Node0: CPU0 memory AAA MB\n\u003e Node1: CPU1 memory AAA MB\n\u003e Node2: CPU2 memory AAA MB\n\u003e Node3: CPU3 memory AAA MB\n\u003e Node4: Memory XXX GB\n\u003e\n\u003e AAA is very small value (below 16MB) and will be omitted by ia64 bootstrap.\n\u003e After boot, only Node 4 has valid memory (but have no cpu.)\n\u003e\n\u003e Maybe this is memory-interleave by firmware config.\n\nChristoph Lameter \u003cclameter@sgi.com\u003e wrote:\n\n\u003e Future SGI platforms (actually also current one can have but nothing like\n\u003e that is deployed to my knowledge) have nodes with only cpus. Current SGI\n\u003e platforms have nodes with just I/O that we so far cannot manage in the\n\u003e core. So the arch code maps them to the nearest memory node.\n\nLee Schermerhorn \u003cLee.Schermerhorn@hp.com\u003e wrote:\n\n\u003e For the HP platforms, we can configure each cell with from 0% to 100%\n\u003e \"cell local memory\". When we configure with \u003c100% CLM, the \"missing\n\u003e percentages\" are interleaved by hardware on a cache-line granularity to\n\u003e improve bandwidth at the expense of latency for numa-challenged\n\u003e applications [and OSes, but not our problem ;-)]. When we boot Linux on\n\u003e such a config, all of the real nodes have no memory--it all resides in a\n\u003e single interleaved pseudo-node.\n\u003e\n\u003e When we boot Linux on a 100% CLM configuration [\u003d\u003d NUMA], we still have\n\u003e the interleaved pseudo-node. It contains a few hundred MB stolen from\n\u003e the real nodes to contain the DMA zone. [Interleaved memory resides at\n\u003e phys addr 0]. The memoryless-nodes patches, along with the zoneorder\n\u003e patches, support this config as well.\n\u003e\n\u003e Also, when we boot a NUMA config with the \"mem\u003d\" command line,\n\u003e specifying less memory than actually exists, Linux takes the excluded\n\u003e memory \"off the top\" rather than distributing it across the nodes. This\n\u003e can result in memoryless nodes, as well.\n\u003e\n\nThis patch:\n\nPreparation for memoryless node patches.\n\nProvide a generic way to keep nodemasks describing various characteristics of\nNUMA nodes.\n\nRemove the node_online_map and the node_possible map and realize the same\nfunctionality using two nodes stats: N_POSSIBLE and N_ONLINE.\n\n[Lee.Schermerhorn@hp.com: Initialize N_*_MEMORY and N_CPU masks for non-NUMA config]\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nTested-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Bob Picco \u003cbob.picco@hp.com\u003e\nCc: Nishanth Aravamudan \u003cnacc@us.ibm.com\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nCc: Mel Gorman \u003cmel@skynet.ie\u003e\nSigned-off-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nCc: \"Serge E. Hallyn\" \u003cserge@hallyn.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "55144768e100b68447f44c5e5c9deb155ad661bd", "tree": "9889682ee3b64627ad9eb504f30672fad844752f", "parents": [ "03158cd7eb3374843de68421142ca5900df845d9" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:25:26 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:58 2007 -0700" }, "message": "fs: remove some AOP_TRUNCATED_PAGE\n\nprepare/commit_write no longer returns AOP_TRUNCATED_PAGE since OCFS2 and\nGFS2 were converted to the new aops, so we can make some simplifications\nfor that.\n\n[michal.k.k.piotrowski@gmail.com: fix warning]\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nCc: Michael Halcrow \u003cmhalcrow@us.ibm.com\u003e\nCc: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\nCc: Steven Whitehouse \u003cswhiteho@redhat.com\u003e\nSigned-off-by: Michal Piotrowski \u003cmichal.k.k.piotrowski@gmail.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "89e107877b65bf6eff1d63a1302dee9a091586f5", "tree": "8a120a04c46f19229d1cf9b9c546f1818cf84c44", "parents": [ "7765ec26ae1c01bb29bedf910e4efcced8cc81d2" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:25:07 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:55 2007 -0700" }, "message": "fs: new cont helpers\n\nRework the generic block \"cont\" routines to handle the new aops. Supporting\ncont_prepare_write would take quite a lot of code to support, so remove it\ninstead (and we later convert all filesystems to use it).\n\nwrite_begin gets passed AOP_FLAG_CONT_EXPAND when called from\ngeneric_cont_expand, so filesystems can avoid the old hacks they used.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nCc: OGAWA Hirofumi \u003chirofumi@mail.parknet.co.jp\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "800d15a53e7d14fa26495b7b95d3bfe7877dd69d", "tree": "09a59e32043f1b601b73ad6105e7634679337884", "parents": [ "674b892ede247ef4fb8d00918f02e29c32b9bbaf" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:25:03 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:55 2007 -0700" }, "message": "implement simple fs aops\n\nImplement new aops for some of the simpler filesystems.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "674b892ede247ef4fb8d00918f02e29c32b9bbaf", "tree": "8e170040e64073932ba559cc31f3ec9802337406", "parents": [ "8268f5a7415d914fc855a86aa2284ac819dc6b2e" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:25:03 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:55 2007 -0700" }, "message": "mm: restore KERNEL_DS optimisations\n\nRestore the KERNEL_DS optimisation, especially helpful to the 2copy write\npath.\n\nThis may be a pretty questionable gain in most cases, especially after the\nlegacy 2copy write path is removed, but it doesn\u0027t cost much.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "afddba49d18f346e5cc2938b6ed7c512db18ca68", "tree": "4726e3d3b0e9e8e5b5d3b2b0cccb36446bbdf3ca", "parents": [ "637aff46f94a754207c80c8c64bf1b74f24b967d" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:25:01 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:55 2007 -0700" }, "message": "fs: introduce write_begin, write_end, and perform_write aops\n\nThese are intended to replace prepare_write and commit_write with more\nflexible alternatives that are also able to avoid the buffered write\ndeadlock problems efficiently (which prepare_write is unable to do).\n\n[mark.fasheh@oracle.com: API design contributions, code review and fixes]\n[akpm@linux-foundation.org: various fixes]\n[dmonakhov@sw.ru: new aop block_write_begin fix]\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\nSigned-off-by: Dmitriy Monakhov \u003cdmonakhov@openvz.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "2f718ffc16c43a435d12919c75dbfad518abd056", "tree": "55588cb2815d844e9d0b2404cf8ceafe98b5c55d", "parents": [ "08291429cfa6258c4cd95d8833beb40f828b194e" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:59 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:55 2007 -0700" }, "message": "mm: buffered write iterator\n\nAdd an iterator data structure to operate over an iovec. Add usercopy\noperators needed by generic_file_buffered_write, and convert that function\nover.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "08291429cfa6258c4cd95d8833beb40f828b194e", "tree": "50a206f0f0e7a5400b44073107ff00517e6f50ac", "parents": [ "4a9e5ef1f4f15205e477817a5cefc34bd3f65f55" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:59 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: fix pagecache write deadlocks\n\nModify the core write() code so that it won\u0027t take a pagefault while holding a\nlock on the pagecache page. There are a number of different deadlocks possible\nif we try to do such a thing:\n\n1. generic_buffered_write\n2. lock_page\n3. prepare_write\n4. unlock_page+vmtruncate\n5. copy_from_user\n6. mmap_sem(r)\n7. handle_mm_fault\n8. lock_page (filemap_nopage)\n9. commit_write\n10. unlock_page\n\na. sys_munmap / sys_mlock / others\nb. mmap_sem(w)\nc. make_pages_present\nd. get_user_pages\ne. handle_mm_fault\nf. lock_page (filemap_nopage)\n\n2,8\t- recursive deadlock if page is same\n2,8;2,8\t- ABBA deadlock is page is different\n2,6;b,f\t- ABBA deadlock if page is same\n\nThe solution is as follows:\n1. If we find the destination page is uptodate, continue as normal, but use\n atomic usercopies which do not take pagefaults and do not zero the uncopied\n tail of the destination. The destination is already uptodate, so we can\n commit_write the full length even if there was a partial copy: it does not\n matter that the tail was not modified, because if it is dirtied and written\n back to disk it will not cause any problems (uptodate *means* that the\n destination page is as new or newer than the copy on disk).\n\n1a. The above requires that fault_in_pages_readable correctly returns access\n information, because atomic usercopies cannot distinguish between\n non-present pages in a readable mapping, from lack of a readable mapping.\n\n2. If we find the destination page is non uptodate, unlock it (this could be\n made slightly more optimal), then allocate a temporary page to copy the\n source data into. Relock the destination page and continue with the copy.\n However, instead of a usercopy (which might take a fault), copy the data\n from the pinned temporary page via the kernel address space.\n\n(also, rename maxlen to seglen, because it was confusing)\n\nThis increases the CPU/memory copy cost by almost 50% on the affected\nworkloads. That will be solved by introducing a new set of pagecache write\naops in a subsequent patch.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "4a9e5ef1f4f15205e477817a5cefc34bd3f65f55", "tree": "51f52086ecbccc3f41955d8d12293ef7a566a05b", "parents": [ "eb2be189317d031895b5ca534fbf735eb546158b" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:58 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: write iovec cleanup\n\nHide some of the open-coded nr_segs tests into the iovec helpers. This is all\nto simplify generic_file_buffered_write, because that gets more complex in the\nnext patch.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "eb2be189317d031895b5ca534fbf735eb546158b", "tree": "8f1eda7af3be7285244a6f1ad77682e90d403c7b", "parents": [ "64649a58919e66ec21792dbb6c48cb3da22cbd7f" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:57 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: buffered write cleanup\n\nQuite a bit of code is used in maintaining these \"cached pages\" that are\nprobably pretty unlikely to get used. It would require a narrow race where\nthe page is inserted concurrently while this process is allocating a page\nin order to create the spare page. Then a multi-page write into an uncached\npart of the file, to make use of it.\n\nNext, the buffered write path (and others) uses its own LRU pagevec when it\nshould be just using the per-CPU LRU pagevec (which will cut down on both data\nand code size cacheline footprint). Also, these private LRU pagevecs are\nemptied after just a very short time, in contrast with the per-CPU pagevecs\nthat are persistent. Net result: 7.3 times fewer lru_lock acquisitions required\nto add the pages to pagecache for a bulk write (in 4K chunks).\n\n[this gets rid of some cond_resched() calls in readahead.c and mpage.c due\n to clashes in -mm. What put them there, and why? ]\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "64649a58919e66ec21792dbb6c48cb3da22cbd7f", "tree": "054605b7da7bad43c299ef66de9d33fda87cd38d", "parents": [ "5fe172370687e03cc6ba8dca990b75db18ff9bb3" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:56 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: trim more holes\n\nIf prepare_write fails with AOP_TRUNCATED_PAGE, or if commit_write fails, then\nwe may have failed the write operation despite prepare_write having\ninstantiated blocks past i_size. Fix this, and consolidate the trimming into\none place.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "5fe172370687e03cc6ba8dca990b75db18ff9bb3", "tree": "8ce000fee15fa9ccd4313018f7f9ab1ffbc8cdaa", "parents": [ "ae37461c70bc8c8416cad1bab13a9898ed030aa1" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:56 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: debug write deadlocks\n\nAllow CONFIG_DEBUG_VM to switch off the prefaulting logic, to simulate the\nMakes the race much easier to hit.\n\nThis is useful for demonstration and testing purposes, but is removed in a\nsubsequent patch.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "ae37461c70bc8c8416cad1bab13a9898ed030aa1", "tree": "7ba24faab6f89651fd2b3d33563847764744d544", "parents": [ "6814d7a91289ceb143285975e244a8f00fd3a830" ], "author": { "name": "Andrew Morton", "email": "akpm@osdl.org", "time": "Tue Oct 16 01:24:55 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: clean up buffered write code\n\nRename some variables and fix some types.\n\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6814d7a91289ceb143285975e244a8f00fd3a830", "tree": "e439e5d802d6ebaa73bac8b08c311ca60d4c8ee4", "parents": [ "4b49643fbb3fa8bf4910f82be02d45e94e8972a4" ], "author": { "name": "Andrew Morton", "email": "akpm@osdl.org", "time": "Tue Oct 16 01:24:54 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "Revert \"[PATCH] generic_file_buffered_write(): deadlock on vectored write\"\n\nThis reverts commit 6527c2bdf1f833cc18e8f42bd97973d583e4aa83, which\nfixed the following bug:\n\n When prefaulting in the pages in generic_file_buffered_write(), we only\n faulted in the pages for the firts segment of the iovec. If the second of\n successive segment described a mmapping of the page into which we\u0027re\n write()ing, and that page is not up-to-date, the fault handler tries to lock\n the already-locked page (to bring it up to date) and deadlocks.\n\n An exploit for this bug is in writev-deadlock-demo.c, in\n http://www.zip.com.au/~akpm/linux/patches/stuff/ext3-tools.tar.gz.\n\n (These demos assume blocksize \u003c PAGE_CACHE_SIZE).\n\nThe problem with this fix is that it takes the kernel back to doing a single\nprepare_write()/commit_write() per iovec segment. So in the worst case we\u0027ll\nrun prepare_write+commit_write 1024 times where we previously would have run\nit once. The other problem with the fix is that it fix all the locking problems.\n\n\u003cinsert numbers obtained via ext3-tools\u0027s writev-speed.c here\u003e\n\nAnd apparently this change killed NFS overwrite performance, because, I\nsuppose, it talks to the server for each prepare_write+commit_write.\n\nSo just back that patch out - we\u0027ll be fixing the deadlock by other means.\n\nNick says: also it only ever actually papered over the bug, because after\nfaulting in the pages, they might be unmapped or reclaimed.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "4b49643fbb3fa8bf4910f82be02d45e94e8972a4", "tree": "ee3bc7881a1e0b71fa093254bb725ad467d8ef72", "parents": [ "41cb8ac025dbbf6782eae10d231e7e2336ad3724" ], "author": { "name": "Andrew Morton", "email": "akpm@osdl.org", "time": "Tue Oct 16 01:24:54 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "Revert \"[PATCH] generic_file_buffered_write(): handle zero-length iovec segments\"\n\nThis reverts commit 81b0c8713385ce1b1b9058e916edcf9561ad76d6, which was\na bugfix against 6527c2bdf1f833cc18e8f42bd97973d583e4aa83 (\"[PATCH]\ngeneric_file_buffered_write(): deadlock on vectored write\"), which we\nalso revert.\n\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "41cb8ac025dbbf6782eae10d231e7e2336ad3724", "tree": "0070e26ee574338a4a3f56f574e1c126a4357c8e", "parents": [ "902aaed0d983dfd459fcb2b678608d4584782200" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:53 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: revert KERNEL_DS buffered write optimisation\n\nRevert the patch from Neil Brown to optimise NFSD writev handling.\n\nCc: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "902aaed0d983dfd459fcb2b678608d4584782200", "tree": "00b18597d87101ba1b37e5dcfde1d91312870160", "parents": [ "754af6f5a85fcd1ecb456851d20c65e4c6ce10ab" ], "author": { "name": "Hisashi Hifumi", "email": "hifumi.hisashi@oss.ntt.co.jp", "time": "Tue Oct 16 01:24:52 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: use pagevec to rotate reclaimable page\n\nWhile running some memory intensive load, system response deteriorated just\nafter swap-out started.\n\nThe cause of this problem is that when a PG_reclaim page is moved to the tail\nof the inactive LRU list in rotate_reclaimable_page(), lru_lock spin lock is\nacquired every page writeback . This deteriorates system performance and\nmakes interrupt hold off time longer when swap-out started.\n\nFollowing patch solves this problem. I use pagevec in rotating reclaimable\npages to mitigate LRU spin lock contention and reduce interrupt hold off time.\n\nI did a test that allocating and touching pages in multiple processes, and\npinging to the test machine in flooding mode to measure response under memory\nintensive load.\n\nThe test result is:\n\n\t-2.6.23-rc5\n\t--- testmachine ping statistics ---\n\t3000 packets transmitted, 3000 received, 0% packet loss, time 53222ms\n\trtt min/avg/max/mdev \u003d 0.074/0.652/172.228/7.176 ms, pipe 11, ipg/ewma\n17.746/0.092 ms\n\n\t-2.6.23-rc5-patched\n\t--- testmachine ping statistics ---\n\t3000 packets transmitted, 3000 received, 0% packet loss, time 51924ms\n\trtt min/avg/max/mdev \u003d 0.072/0.108/3.884/0.114 ms, pipe 2, ipg/ewma\n17.314/0.091 ms\n\nMax round-trip-time was improved.\n\nThe test machine spec is that 4CPU(3.16GHz, Hyper-threading enabled)\n8GB memory , 8GB swap.\n\nI did ping test again to observe performance deterioration caused by taking\na ref.\n\n\t-2.6.23-rc6-with-modifiedpatch\n\t--- testmachine ping statistics ---\n\t3000 packets transmitted, 3000 received, 0% packet loss, time 53386ms\n\trtt min/avg/max/mdev \u003d 0.074/0.110/4.716/0.147 ms, pipe 2, ipg/ewma 17.801/0.129 ms\n\nThe result for my original patch is as follows.\n\n\t-2.6.23-rc5-with-originalpatch\n\t--- testmachine ping statistics ---\n\t3000 packets transmitted, 3000 received, 0% packet loss, time 51924ms\n\trtt min/avg/max/mdev \u003d 0.072/0.108/3.884/0.114 ms, pipe 2, ipg/ewma 17.314/0.091 ms\n\nThe influence to response was small.\n\n[akpm@linux-foundation.org: fix uninitalised var warning]\n[hugh@veritas.com: fix locking]\n[randy.dunlap@oracle.com: fix function declaration]\n[hugh@veritas.com: fix BUG at include/linux/mm.h:220!]\n[hugh@veritas.com: kill redundancy in rotate_reclaimable_page]\n[hugh@veritas.com: move_tail_pages into lru_add_drain]\nSigned-off-by: Hisashi Hifumi \u003chifumi.hisashi@oss.ntt.co.jp\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "754af6f5a85fcd1ecb456851d20c65e4c6ce10ab", "tree": "8c985bfd704a8c993d6ca992725969c6fc5c9e5a", "parents": [ "32a4330d4156e55a4888a201f484dbafed9504ed" ], "author": { "name": "Lee Schermerhorn", "email": "Lee.Schermerhorn@hp.com", "time": "Tue Oct 16 01:24:51 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "Mem Policy: add MPOL_F_MEMS_ALLOWED get_mempolicy() flag\n\nAllow an application to query the memories allowed by its context.\n\nUpdated numa_memory_policy.txt to mention that applications can use this to\nobtain allowed memories for constructing valid policies.\n\nTODO: update out-of-tree libnuma wrapper[s], or maybe add a new\nwrapper--e.g., numa_get_mems_allowed() ?\n\nAlso, update numa syscall man pages.\n\nTested with memtoy V\u003e\u003d0.13.\n\nSigned-off-by: Lee Schermerhorn \u003clee.schermerhorn@hp.com\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: Andi Kleen \u003cak@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "32a4330d4156e55a4888a201f484dbafed9504ed", "tree": "1339473dab0fe493dc9feb59833cdf23cf2b1780", "parents": [ "8691f3a72f32f8b3ed535faa27140b3ae293c90b" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Tue Oct 16 01:24:50 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: prevent kswapd from freeing excessive amounts of lowmem\n\nThe current VM can get itself into trouble fairly easily on systems with a\nsmall ZONE_HIGHMEM, which is common on i686 computers with 1GB of memory.\n\nOn one side, page_alloc() will allocate down to zone-\u003epages_low, while on\nthe other side, kswapd() and balance_pgdat() will try to free memory from\nevery zone, until every zone has more free pages than zone-\u003epages_high.\n\nHighmem can be filled up to zone-\u003epages_low with page tables, ramfs,\nvmalloc allocations and other unswappable things quite easily and without\nmany bad side effects, since we still have a huge ZONE_NORMAL to do future\nallocations from.\n\nHowever, as long as the number of free pages in the highmem zone is below\nzone-\u003epages_high, kswapd will continue swapping things out from\nZONE_NORMAL, too!\n\nSami Farin managed to get his system into a stage where kswapd had freed\nabout 700MB of low memory and was still \"going strong\".\n\nThe attached patch will make kswapd stop paging out data from zones when\nthere is more than enough memory free. We do go above zone-\u003epages_high in\norder to keep pressure between zones equal in normal circumstances, but the\npatch should prevent the kind of excesses that made Sami\u0027s computer totally\nunusable.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "8691f3a72f32f8b3ed535faa27140b3ae293c90b", "tree": "78ed23998e99de2fb4fb5b135c1fa4ad70612ef0", "parents": [ "26fb1589cb0aaec3a0b4418c54f30c1a2b1781f6" ], "author": { "name": "Jesper Juhl", "email": "jesper.juhl@gmail.com", "time": "Tue Oct 16 01:24:49 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:54 2007 -0700" }, "message": "mm: no need to cast vmalloc() return value in zone_wait_table_init()\n\nvmalloc() returns a void pointer, so there\u0027s no need to cast its\nreturn value in mm/page_alloc.c::zone_wait_table_init().\n\nSigned-off-by: Jesper Juhl \u003cjesper.juhl@gmail.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "ef8b4520bd9f8294ffce9abd6158085bde5dc902", "tree": "c099a16691ac06208f4d3d65b71e7adaf7361fcd", "parents": [ "0da7e01f5f37f441cccd7c8c0586e06db0981907" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:24:46 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "Slab allocators: fail if ksize is called with a NULL parameter\n\nA NULL pointer means that the object was not allocated. One cannot\ndetermine the size of an object that has not been allocated. Currently we\nreturn 0 but we really should BUG() on attempts to determine the size of\nsomething nonexistent.\n\nkrealloc() interprets NULL to mean a zero sized object. Handle that\nseparately in krealloc().\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nAcked-by: Pekka Enberg \u003cpenberg@cs.helsinki.fi\u003e\nCc: Matt Mackall \u003cmpm@selenic.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "0da7e01f5f37f441cccd7c8c0586e06db0981907", "tree": "3d43edc4166667948415de84254bd60cfb457734", "parents": [ "2408c55037c3f7d51a8a100025c47595e71b838c" ], "author": { "name": "Dean Nelson", "email": "dcn@sgi.com", "time": "Tue Oct 16 01:24:45 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "calculation of pgoff in do_linear_fault() uses mixed units\n\nThe calculation of pgoff in do_linear_fault() should use PAGE_SHIFT and not\nPAGE_CACHE_SHIFT since vma-\u003evm_pgoff is in units of PAGE_SIZE and not\nPAGE_CACHE_SIZE. At the moment linux/pagemap.h has PAGE_CACHE_SHIFT\ndefined as PAGE_SHIFT, but should that ever change this calculation would\nbreak.\n\nSigned-off-by: Dean Nelson \u003cdcn@sgi.com\u003e\nAcked-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "2408c55037c3f7d51a8a100025c47595e71b838c", "tree": "71940f72dbadf6a6744ffbd062ef4fd8754aa623", "parents": [ "c92ff1bde06f69d59b40f3194016aee51cc5da55" ], "author": { "name": "Satyam Sharma", "email": "satyam@infradead.org", "time": "Tue Oct 16 01:24:44 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "{slub, slob}: use unlikely() for kfree(ZERO_OR_NULL_PTR) check\n\nConsidering kfree(NULL) would normally occur only in error paths and\nkfree(ZERO_SIZE_PTR) is uncommon as well, so let\u0027s use unlikely() for the\ncondition check in SLUB\u0027s and SLOB\u0027s kfree() to optimize for the common\ncase. SLAB has this already.\n\nSigned-off-by: Satyam Sharma \u003csatyam@infradead.org\u003e\nCc: Pekka Enberg \u003cpenberg@cs.helsinki.fi\u003e\nCc: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "b55ed816235cf41c29159d22a4cdeec7deb5821c", "tree": "73245c1da2c5352c526d4d25f02e69943b4b0b15", "parents": [ "45726cb43d11b288c58243a26010f397054222f0" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:42 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "mm: clarify __add_to_swap_cache locking\n\n__add_to_swap_cache unconditionally sets the page locked, which can be a bit\nalarming to the unsuspecting reader: in the code paths where the page is\nvisible to other CPUs, the page should be (and is) already locked.\n\nInstead, just add a check to ensure the page is locked here, and teach the one\npath relying on the old behaviour to call SetPageLocked itself.\n\n[hugh@veritas.com: locking fix]\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Hugh Dickins \u003chugh@veritas.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "45726cb43d11b288c58243a26010f397054222f0", "tree": "ae1804a34f314be18865add063d660cf783e45be", "parents": [ "001281881067a5998384c6669bc8dbbbab8456c4" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:41 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "mm: improve find_lock_page\n\nfind_lock_page does not need to recheck -\u003eindex because if the page is in the\nright mapping then the index must be the same. Also, tree_lock does not need\nto be retaken after the page is locked in order to test that -\u003emapping has not\nchanged, because holding the page lock pins its mapping.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "001281881067a5998384c6669bc8dbbbab8456c4", "tree": "6229d16c737782a36795354b218c74d4d2749752", "parents": [ "557ed1fa2620dc119adb86b34c614e152a629a80" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:40 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "mm: use lockless radix-tree probe\n\nProbing pages and radix_tree_tagged are lockless operations with the lockless\nradix-tree. Convert these users to RCU locking rather than using tree_lock.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "557ed1fa2620dc119adb86b34c614e152a629a80", "tree": "d00b31a7f197583c2bd8fffa1fd135fbbb5d6abc", "parents": [ "aadb4bc4a1f9108c1d0fbd121827c936c2ed4217" ], "author": { "name": "Nick Piggin", "email": "npiggin@suse.de", "time": "Tue Oct 16 01:24:40 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "remove ZERO_PAGE\n\nThe commit b5810039a54e5babf428e9a1e89fc1940fabff11 contains the note\n\n A last caveat: the ZERO_PAGE is now refcounted and managed with rmap\n (and thus mapcounted and count towards shared rss). These writes to\n the struct page could cause excessive cacheline bouncing on big\n systems. There are a number of ways this could be addressed if it is\n an issue.\n\nAnd indeed this cacheline bouncing has shown up on large SGI systems.\nThere was a situation where an Altix system was essentially livelocked\ntearing down ZERO_PAGE pagetables when an HPC app aborted during startup.\nThis situation can be avoided in userspace, but it does highlight the\npotential scalability problem with refcounting ZERO_PAGE, and corner\ncases where it can really hurt (we don\u0027t want the system to livelock!).\n\nThere are several broad ways to fix this problem:\n1. add back some special casing to avoid refcounting ZERO_PAGE\n2. per-node or per-cpu ZERO_PAGES\n3. remove the ZERO_PAGE completely\n\nI will argue for 3. The others should also fix the problem, but they\nresult in more complex code than does 3, with little or no real benefit\nthat I can see.\n\nWhy? Inserting a ZERO_PAGE for anonymous read faults appears to be a\nfalse optimisation: if an application is performance critical, it would\nnot be doing many read faults of new memory, or at least it could be\nexpected to write to that memory soon afterwards. If cache or memory use\nis critical, it should not be working with a significant number of\nZERO_PAGEs anyway (a more compact representation of zeroes should be\nused).\n\nAs a sanity check -- mesuring on my desktop system, there are never many\nmappings to the ZERO_PAGE (eg. 2 or 3), thus memory usage here should not\nincrease much without it.\n\nWhen running a make -j4 kernel compile on my dual core system, there are\nabout 1,000 mappings to the ZERO_PAGE created per second, but about 1,000\nZERO_PAGE COW faults per second (less than 1 ZERO_PAGE mapping per second\nis torn down without being COWed). So removing ZERO_PAGE will save 1,000\npage faults per second when running kbuild, while keeping it only saves\nless than 1 page clearing operation per second. 1 page clear is cheaper\nthan a thousand faults, presumably, so there isn\u0027t an obvious loss.\n\nNeither the logical argument nor these basic tests give a guarantee of no\nregressions. However, this is a reasonable opportunity to try to remove\nthe ZERO_PAGE from the pagefault path. If it is found to cause regressions,\nwe can reintroduce it and just avoid refcounting it.\n\nThe /dev/zero ZERO_PAGE usage and TLB tricks also get nuked. I don\u0027t see\nmuch use to them except on benchmarks. All other users of ZERO_PAGE are\nconverted just to use ZERO_PAGE(0) for simplicity. We can look at\nreplacing them all and maybe ripping out ZERO_PAGE completely when we are\nmore satisfied with this solution.\n\nSigned-off-by: Nick Piggin \u003cnpiggin@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus \"snif\" Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "aadb4bc4a1f9108c1d0fbd121827c936c2ed4217", "tree": "879b7c9ba11a65958e4477c563602e08d9e6635f", "parents": [ "57f6b96c09c30e444e0d3fc3080feba037657a7b" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:24:38 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "SLUB: direct pass through of page size or higher kmalloc requests\n\nThis gets rid of all kmalloc caches larger than page size. A kmalloc\nrequest larger than PAGE_SIZE \u003e 2 is going to be passed through to the page\nallocator. This works both inline where we will call __get_free_pages\ninstead of kmem_cache_alloc and in __kmalloc.\n\nkfree is modified to check if the object is in a slab page. If not then\nthe page is freed via the page allocator instead. Roughly similar to what\nSLOB does.\n\nAdvantages:\n- Reduces memory overhead for kmalloc array\n- Large kmalloc operations are faster since they do not\n need to pass through the slab allocator to get to the\n page allocator.\n- Performance increase of 10%-20% on alloc and 50% on free for\n PAGE_SIZEd allocations.\n SLUB must call page allocator for each alloc anyways since\n the higher order pages which that allowed avoiding the page alloc calls\n are not available in a reliable way anymore. So we are basically removing\n useless slab allocator overhead.\n- Large kmallocs yields page aligned object which is what\n SLAB did. Bad things like using page sized kmalloc allocations to\n stand in for page allocate allocs can be transparently handled and are not\n distinguishable from page allocator uses.\n- Checking for too large objects can be removed since\n it is done by the page allocator.\n\nDrawbacks:\n- No accounting for large kmalloc slab allocations anymore\n- No debugging of large kmalloc slab allocations.\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "57f6b96c09c30e444e0d3fc3080feba037657a7b", "tree": "4b62e2f63352b74d6c9dd1ed1142b971dcd207e3", "parents": [ "b2c3843b1e25e2c67347c4671f33fbe6f5067e6b" ], "author": { "name": "Fengguang Wu", "email": "wfg@mail.ustc.edu.cn", "time": "Tue Oct 16 01:24:37 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "filemap: convert some unsigned long to pgoff_t\n\nConvert some \u0027unsigned long\u0027 to pgoff_t.\n\nSigned-off-by: Fengguang Wu \u003cwfg@mail.ustc.edu.cn\u003e\nCc: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "b2c3843b1e25e2c67347c4671f33fbe6f5067e6b", "tree": "e1b161e5c7169baab8f71615879eb3b46ebe9a9e", "parents": [ "f2e189827a914b66e435e68b1c9e37775cb995ed" ], "author": { "name": "Fengguang Wu", "email": "wfg@mail.ustc.edu.cn", "time": "Tue Oct 16 01:24:37 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:53 2007 -0700" }, "message": "filemap: trivial code cleanups\n\n- remove unused local next_index in do_generic_mapping_read()\n- remove a redudant page_cache_read() declaration\n\nSigned-off-by: Fengguang Wu \u003cwfg@mail.ustc.edu.cn\u003e\nCc: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "535443f51543df61111bbd234300ae549d220448", "tree": "e9b022969776132b9fd3bde351673edc006a1bac", "parents": [ "7ff81078d8b9f3d05a27b7bd3786ffb1ef1b0d1f" ], "author": { "name": "Fengguang Wu", "email": "wfg@mail.ustc.edu.cn", "time": "Tue Oct 16 01:24:36 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:52 2007 -0700" }, "message": "readahead: remove several readahead macros\n\nRemove VM_MAX_CACHE_HIT, MAX_RA_PAGES and MIN_RA_PAGES.\n\nSigned-off-by: Fengguang Wu \u003cwfg@mail.ustc.edu.cn\u003e\nCc: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "7ff81078d8b9f3d05a27b7bd3786ffb1ef1b0d1f", "tree": "ac73cf0c8325783a28c4d16c783f6fd96d17be7c", "parents": [ "6b10c6c9fbfe754e8482efb8c8b84f8e40c0f2eb" ], "author": { "name": "Fengguang Wu", "email": "wfg@mail.ustc.edu.cn", "time": "Tue Oct 16 01:24:35 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:52 2007 -0700" }, "message": "readahead: remove the local copy of ra in do_generic_mapping_read()\n\nThe local copy of ra in do_generic_mapping_read() can now go away.\n\nIt predates readanead(req_size). In a time when the readahead code was called\non *every* single page. Hence a local has to be made to reduce the chance of\nthe readahead state being overwritten by a concurrent reader. More details\nin: Linux: Random File I/O Regressions In 2.6\n\u003chttp://kerneltrap.org/node/3039\u003e\n\nCc: Nick Piggin \u003cnickpiggin@yahoo.com.au\u003e\nSigned-off-by: Fengguang Wu \u003cwfg@mail.ustc.edu.cn\u003e\nCc: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6b10c6c9fbfe754e8482efb8c8b84f8e40c0f2eb", "tree": "08f275b1e8d2e9c93bb46367611c43ab88f8f8dc", "parents": [ "6df8ba4f8a4c4abca9ccad10441d0dddbdff301c" ], "author": { "name": "Fengguang Wu", "email": "wfg@mail.ustc.edu.cn", "time": "Tue Oct 16 01:24:34 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:52 2007 -0700" }, "message": "readahead: basic support of interleaved reads\n\nThis is a simplified version of the pagecache context based readahead. It\nhandles the case of multiple threads reading on the same fd and invalidating\neach others\u0027 readahead state. It does the trick by scanning the pagecache and\nrecovering the current read stream\u0027s readahead status.\n\nThe algorithm works in a opportunistic way, in that it does not try to detect\ninterleaved reads _actively_, which requires a probe into the page cache\n(which means a little more overhead for random reads). It only tries to\nhandle a previously started sequential readahead whose state was overwritten\nby another concurrent stream, and it can do this job pretty well.\n\nNegative and positive examples(or what you can expect from it):\n\n1) it cannot detect and serve perfect request-by-request interleaved reads\n right:\n\ttime\tstream 1 stream 2\n\t0 \t1\n\t1 \t 1001\n\t2 \t2\n\t3 \t 1002\n\t4 \t3\n\t5 \t 1003\n\t6 \t4\n\t7 \t 1004\n\t8 \t5\n\t9\t 1005\n\nHere no single readahead will be carried out.\n\n2) However, if it\u0027s two concurrent reads by two threads, the chance of the\n initial sequential readahead be started is huge. Once the first sequential\n readahead is started for a stream, this patch will ensure that the readahead\n window continues to rampup and won\u0027t be disturbed by other streams.\n\n\ttime\tstream 1 stream 2\n\t0 \t1\n\t1 \t2\n\t2 \t 1001\n\t3 \t3\n\t4 \t 1002\n\t5 \t 1003\n\t6 \t4\n\t7 \t5\n\t8 \t 1004\n\t9 \t6\n\t10\t 1005\n\t11\t7\n\t12\t 1006\n\t13\t 1007\n\nHere stream 1 will start a readahead at page 2, and stream 2 will start its\nfirst readahead at page 1003. From then on the two streams will be served\nright.\n\nCc: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Fengguang Wu \u003cwfg@mail.ustc.edu.cn\u003e\nCc: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "f4e6b498d6e06742d72706ef50593a9c4dd72214", "tree": "74a573302b2ea086c0d21907175be604f110f5b1", "parents": [ "0bb7ba6b9c358c12084a3cbc6ac08c8d1e973937" ], "author": { "name": "Fengguang Wu", "email": "wfg@mail.ustc.edu.cn", "time": "Tue Oct 16 01:24:33 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:52 2007 -0700" }, "message": "readahead: combine file_ra_state.prev_index/prev_offset into prev_pos\n\nCombine the file_ra_state members\n\t\t\t\tunsigned long prev_index\n\t\t\t\tunsigned int prev_offset\ninto\n\t\t\t\tloff_t prev_pos\n\nIt is more consistent and better supports huge files.\n\nThanks to Peter for the nice proposal!\n\n[akpm@linux-foundation.org: fix shift overflow]\nCc: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Fengguang Wu \u003cwfg@mail.ustc.edu.cn\u003e\nCc: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "0bb7ba6b9c358c12084a3cbc6ac08c8d1e973937", "tree": "81dc745c660557af4d9ea8b9816846d0395ece20", "parents": [ "937085aa35cc873d427d250a1e304d641af24628" ], "author": { "name": "Fengguang Wu", "email": "wfg@mail.ustc.edu.cn", "time": "Tue Oct 16 01:24:32 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:52 2007 -0700" }, "message": "readahead: mmap read-around simplification\n\nFold file_ra_state.mmap_hit into file_ra_state.mmap_miss and make it an int.\n\nSigned-off-by: Fengguang Wu \u003cwfg@mail.ustc.edu.cn\u003e\nCc: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "937085aa35cc873d427d250a1e304d641af24628", "tree": "99b0d2d9e6721a72e65b4cb5ad0fc30ec04f4aa2", "parents": [ "43fac94dd62667c83dd2daa5b7ac548512af780a" ], "author": { "name": "Fengguang Wu", "email": "wfg@mail.ustc.edu.cn", "time": "Tue Oct 16 01:24:31 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:52 2007 -0700" }, "message": "readahead: compacting file_ra_state\n\nUse \u0027unsigned int\u0027 instead of \u0027unsigned long\u0027 for readahead sizes.\n\nThis helps reduce memory consumption on 64bit CPU when a lot of files are\nopened.\n\nCC: Andi Kleen \u003candi@firstfloor.org\u003e\nSigned-off-by: Fengguang Wu \u003cwfg@mail.ustc.edu.cn\u003e\nCc: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "43fac94dd62667c83dd2daa5b7ac548512af780a", "tree": "68b8cf73959afd24410f3f398bda5953c7dcbadd", "parents": [ "39e91e433169bdfd5a312654e5988986662afd7f" ], "author": { "name": "Jesper Juhl", "email": "jesper.juhl@gmail.com", "time": "Tue Oct 16 01:24:30 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:52 2007 -0700" }, "message": "Clean up duplicate includes in mm/\n\nThis patch cleans up duplicate includes in\n\tmm/\n\nSigned-off-by: Jesper Juhl \u003cjesper.juhl@gmail.com\u003e\nAcked-by: Paul Mundt \u003clethal@linux-sh.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "1cd7daa51baf0934882bf24a7ae378ec75df665e", "tree": "496ea3e64fc65d01e37f5013b44354ceb8ed3712", "parents": [ "d29eff7bca60c9ee401d691d4562a4abca8de543" ], "author": { "name": "Adrian Bunk", "email": "bunk@stusta.de", "time": "Tue Oct 16 01:24:18 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:51 2007 -0700" }, "message": "slub.c:early_kmem_cache_node_alloc() shouldn\u0027t be __init\n\nWARNING: mm/built-in.o(.text+0x24bd3): Section mismatch: reference to .init.text:early_kmem_cache_node_alloc (between \u0027init_kmem_cache_nodes\u0027 and \u0027calculate_sizes\u0027)\n...\n\nSigned-off-by: Adrian Bunk \u003cbunk@stusta.de\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "29c71111d0557385328211b130246a90f9223b46", "tree": "5588a49ee548d38e15bd7541cec29e069b9e457c", "parents": [ "8f6aac419bd590f535fb110875a51f7db2b62b5b" ], "author": { "name": "Andy Whitcroft", "email": "apw@shadowen.org", "time": "Tue Oct 16 01:24:14 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:51 2007 -0700" }, "message": "vmemmap: generify initialisation via helpers\n\nConvert the common vmemmap population into initialisation helpers for use by\narchitecture vmemmap populators. All architecture implementing the\nSPARSEMEM_VMEMMAP variant supply an architecture specific vmemmap_populate()\ninitialiser, which may make use of the helpers.\n\nThis allows us to clean up and remove the initialisation Kconfig entries.\nWith this patch there is a single SPARSEMEM_VMEMMAP_ENABLE Kconfig option to\nindicate use of that variant.\n\nSigned-off-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "8f6aac419bd590f535fb110875a51f7db2b62b5b", "tree": "64e73e9f7a4b5a68648a2b4b16e66307c3d8d3cf", "parents": [ "540557b9439ec19668553830c90222f9fb0c2e95" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Tue Oct 16 01:24:13 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:51 2007 -0700" }, "message": "Generic Virtual Memmap support for SPARSEMEM\n\nSPARSEMEM is a pretty nice framework that unifies quite a bit of code over all\nthe arches. It would be great if it could be the default so that we can get\nrid of various forms of DISCONTIG and other variations on memory maps. So far\nwhat has hindered this are the additional lookups that SPARSEMEM introduces\nfor virt_to_page and page_address. This goes so far that the code to do this\nhas to be kept in a separate function and cannot be used inline.\n\nThis patch introduces a virtual memmap mode for SPARSEMEM, in which the memmap\nis mapped into a virtually contigious area, only the active sections are\nphysically backed. This allows virt_to_page page_address and cohorts become\nsimple shift/add operations. No page flag fields, no table lookups, nothing\ninvolving memory is required.\n\nThe two key operations pfn_to_page and page_to_page become:\n\n #define __pfn_to_page(pfn) (vmemmap + (pfn))\n #define __page_to_pfn(page) ((page) - vmemmap)\n\nBy having a virtual mapping for the memmap we allow simple access without\nwasting physical memory. As kernel memory is typically already mapped 1:1\nthis introduces no additional overhead. The virtual mapping must be big\nenough to allow a struct page to be allocated and mapped for all valid\nphysical pages. This vill make a virtual memmap difficult to use on 32 bit\nplatforms that support 36 address bits.\n\nHowever, if there is enough virtual space available and the arch already maps\nits 1-1 kernel space using TLBs (f.e. true of IA64 and x86_64) then this\ntechnique makes SPARSEMEM lookups even more efficient than CONFIG_FLATMEM.\nFLATMEM needs to read the contents of the mem_map variable to get the start of\nthe memmap and then add the offset to the required entry. vmemmap is a\nconstant to which we can simply add the offset.\n\nThis patch has the potential to allow us to make SPARSMEM the default (and\neven the only) option for most systems. It should be optimal on UP, SMP and\nNUMA on most platforms. Then we may even be able to remove the other memory\nmodels: FLATMEM, DISCONTIG etc.\n\n[apw@shadowen.org: config cleanups, resplit code etc]\n[kamezawa.hiroyu@jp.fujitsu.com: Fix sparsemem_vmemmap init]\n[apw@shadowen.org: vmemmap: remove excess debugging]\n[apw@shadowen.org: simplify initialisation code and reduce duplication]\n[apw@shadowen.org: pull out the vmemmap code into its own file]\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nCc: \"Luck, Tony\" \u003ctony.luck@intel.com\u003e\nCc: Andi Kleen \u003cak@suse.de\u003e\nCc: \"David S. Miller\" \u003cdavem@davemloft.net\u003e\nCc: Paul Mackerras \u003cpaulus@samba.org\u003e\nCc: Benjamin Herrenschmidt \u003cbenh@kernel.crashing.org\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "540557b9439ec19668553830c90222f9fb0c2e95", "tree": "07dfa0e88580d4101dbb11ebc59348233e18b2f0", "parents": [ "cd881a6b22902b356cacf8fd2e4e895871068eec" ], "author": { "name": "Andy Whitcroft", "email": "apw@shadowen.org", "time": "Tue Oct 16 01:24:11 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 16 09:42:51 2007 -0700" }, "message": "sparsemem: record when a section has a valid mem_map\n\nWe have flags to indicate whether a section actually has a valid mem_map\nassociated with it. This is never set and we rely solely on the present bit\nto indicate a section is valid. By definition a section is not valid if it\nhas no mem_map and there is a window during init where the present bit is set\nbut there is no mem_map, during which pfn_valid() will return true\nincorrectly.\n\nUse the existing SECTION_HAS_MEM_MAP flag to indicate the presence of a valid\nmem_map. Switch valid_section{,_nr} and pfn_valid() to this bit. Add a new\npresent_section{,_nr} and pfn_present() interfaces for those users who care to\nknow that a section is going to be valid.\n\n[akpm@linux-foundation.org: coding-syle fixes]\nSigned-off-by: Andy Whitcroft \u003capw@shadowen.org\u003e\nAcked-by: Mel Gorman \u003cmel@csn.ul.ie\u003e\nCc: Christoph Lameter \u003cclameter@sgi.com\u003e\nCc: \"Luck, Tony\" \u003ctony.luck@intel.com\u003e\nCc: Andi Kleen \u003cak@suse.de\u003e\nCc: \"David S. Miller\" \u003cdavem@davemloft.net\u003e\nCc: Paul Mackerras \u003cpaulus@samba.org\u003e\nCc: Benjamin Herrenschmidt \u003cbenh@kernel.crashing.org\u003e\nCc: KAMEZAWA Hiroyuki \u003ckamezawa.hiroyu@jp.fujitsu.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" } ], "next": "cd881a6b22902b356cacf8fd2e4e895871068eec" }