)]}' { "log": [ { "commit": "9499f5e7ed5224c40706f0cec6542a9916bc7606", "tree": "3e4e1b36d3d549ea356e88e6e44359a887c6ee01", "parents": [ "ef688e151c00e5d529703be9a04fd506df8bc54e" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Fri Jun 12 22:16:35 2009 -0600" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Fri Jun 12 22:16:36 2009 +0930" }, "message": "virtio: add names to virtqueue struct, mapping from devices to queues.\n\nAdd a linked list of all virtqueues for a virtio device: this helps for\ndebugging and is also needed for upcoming interface change.\n\nAlso, add a \"name\" field for clearer debug messages.\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\n" }, { "commit": "20f77f5654042cf484d8964b618faf9d620f639b", "tree": "d5007f2e5ec72062532b7930ef0a6f68708d9658", "parents": [ "8ebf975608aaebd7feb33d77f07ba21a6380e086" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Fri Jun 12 22:16:33 2009 -0600" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Fri Jun 12 22:16:35 2009 +0930" }, "message": "virtio: fix obsolete documentation on probe function\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\n" }, { "commit": "c45a6816c19dee67b8f725e6646d428901a6dc24", "tree": "096e3263fd14e140685bcc3082394ff15f5aeddb", "parents": [ "72e61eb40b55dd57031ec5971e810649f82b0259" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Fri May 02 21:50:50 2008 -0500" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Fri May 02 21:50:50 2008 +1000" }, "message": "virtio: explicit advertisement of driver features\n\nA recent proposed feature addition to the virtio block driver revealed\nsome flaws in the API: in particular, we assume that feature\nnegotiation is complete once a driver\u0027s probe function returns.\n\nThere is nothing in the API to require this, however, and even I\ndidn\u0027t notice when it was violated.\n\nSo instead, we require the driver to specify what features it supports\nin a table, we can then move the feature negotiation into the virtio\ncore. The intersection of device and driver features are presented in\na new \u0027features\u0027 bitmap in the struct virtio_device.\n\nNote that this highlights the difference between Linux unsigned-long\nbitmaps where each unsigned long is in native endian, and a\nstraight-forward little-endian array of bytes.\n\nDrivers can still remove feature bits in their probe routine if they\nreally have to.\n\nAPI changes:\n- dev-\u003econfig-\u003efeature() no longer gets and acks a feature.\n- drivers should advertise their features in the \u0027feature_table\u0027 field\n- use virtio_has_feature() for extra sanity when checking feature bits\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\n" }, { "commit": "2557a933b795c1988c721ebb871cd735128bb9cb", "tree": "96581bd5acac44bbe0399aa445197509d9ae60ff", "parents": [ "4cac04dd63fa3b202ee313ed1afbbd135ab887ee" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Mon Apr 07 14:30:28 2008 +1000" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Apr 07 13:14:22 2008 -0700" }, "message": "virtio: remove overzealous BUG_ON.\n\nThe \u0027disable_cb\u0027 callback is designed as an optimization to tell the host\nwe don\u0027t need callbacks now. As it is not reliable, the debug check is\noverzealous: it can happen on two CPUs at the same time. Document this.\n\nEven if it were reliable, the virtio_net driver doesn\u0027t disable\ncallbacks on transmit so the START_USE/END_USE debugging reentrance\nprotection can be easily tripped even on UP.\n\nThanks to Balaji Rao for the bug report and testing.\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nCC: Balaji Rao \u003cbalajirrao@gmail.com\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "4265f161b6bb7b31163671329b1142b9023bf4e3", "tree": "1494768aaf38407ca8f1336f6ae80301b27135bb", "parents": [ "da74e89d40995600b3b07ac500084920247687ce" ], "author": { "name": "Christian Borntraeger", "email": "borntraeger@de.ibm.com", "time": "Fri Mar 14 14:17:05 2008 +0100" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Mon Mar 17 22:58:21 2008 +1100" }, "message": "virtio: fix race in enable_cb\n\nThere is a race in virtio_net, dealing with disabling/enabling the callback.\nI saw the following oops:\n\nkernel BUG at /space/kvm/drivers/virtio/virtio_ring.c:218!\nillegal operation: 0001 [#1] SMP\nModules linked in: sunrpc dm_mod\nCPU: 2 Not tainted 2.6.25-rc1zlive-host-10623-gd358142-dirty #99\nProcess swapper (pid: 0, task: 000000000f85a610, ksp: 000000000f873c60)\nKrnl PSW : 0404300180000000 00000000002b81a6 (vring_disable_cb+0x16/0x20)\n R:0 T:1 IO:0 EX:0 Key:0 M:1 W:0 P:0 AS:0 CC:3 PM:0 EA:3\nKrnl GPRS: 0000000000000001 0000000000000001 0000000010005800 0000000000000001\n 000000000f3a0900 000000000f85a610 0000000000000000 0000000000000000\n 0000000000000000 000000000f870000 0000000000000000 0000000000001237\n 000000000f3a0920 000000000010ff74 00000000002846f6 000000000fa0bcd8\nKrnl Code: 00000000002b819a: a7110001 tmll %r1,1\n 00000000002b819e: a7840004 brc 8,2b81a6\n 00000000002b81a2: a7f40001 brc 15,2b81a4\n \u003e00000000002b81a6: a51b0001 oill %r1,1\n 00000000002b81aa: 40102000 sth %r1,0(%r2)\n 00000000002b81ae: 07fe bcr 15,%r14\n 00000000002b81b0: eb7ff0380024 stmg %r7,%r15,56(%r15)\n 00000000002b81b6: a7f13e00 tmll %r15,15872\nCall Trace:\n([\u003c000000000fa0bcd0\u003e] 0xfa0bcd0)\n [\u003c00000000002b8350\u003e] vring_interrupt+0x5c/0x6c\n [\u003c000000000010ab08\u003e] do_extint+0xb8/0xf0\n [\u003c0000000000110716\u003e] ext_no_vtime+0x16/0x1a\n [\u003c0000000000107e72\u003e] cpu_idle+0x1c2/0x1e0\n\nThe problem can be triggered with a high amount of host-\u003eguest traffic.\nI think its the following race:\n\npoll says netif_rx_complete\npoll calls enable_cb\nenable_cb opens the interrupt mask\na new packet comes, an interrupt is triggered----\\\nenable_cb sees that there is more work |\nenable_cb disables the interrupt |\n . V\n . interrupt is delivered\n . skb_recv_done does atomic napi test, ok\n some waiting disable_cb is called-\u003echeck fails-\u003ebang!\n .\npoll would do napi check\npoll would do disable_cb\n\nThe fix is to let enable_cb not disable the interrupt again, but expect the\ncaller to do the cleanup if it returns false. In that case, the interrupt is\nonly disabled, if the napi test_set_bit was successful.\n\nSigned-off-by: Christian Borntraeger \u003cborntraeger@de.ibm.com\u003e\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e (cleaned up doco)\n" }, { "commit": "6e5aa7efb27aec7e55b6463fa2c8db594c4226fa", "tree": "060a955e711ac224136157a5410e88dcdab965af", "parents": [ "b3369c1fb410fddeb38a404316c861395f6d6ae8" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Mon Feb 04 23:50:03 2008 -0500" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Mon Feb 04 23:50:03 2008 +1100" }, "message": "virtio: reset function\n\nA reset function solves three problems:\n\n1) It allows us to renegotiate features, eg. if we want to upgrade a\n guest driver without rebooting the guest.\n\n2) It gives us a clean way of shutting down virtqueues: after a reset,\n we know that the buffers won\u0027t be used by the host, and\n\n3) It helps the guest recover from messed-up drivers.\n\nSo we remove the -\u003eshutdown hook, and the only way we now remove\nfeature bits is via reset.\n\nWe leave it to the driver to do the reset before it deletes queues:\nthe balloon driver, for example, needs to chat to the host in its\nremove function.\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\n" }, { "commit": "f957d1f05a1a20bc3b954877c6562a4d53d58bde", "tree": "4eb230a96bb15b74050005f40a7ec383e26d13ad", "parents": [ "18445c4d501b9ab4336f66ef46b092661ddaf336" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Mon Feb 04 23:49:58 2008 -0500" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Mon Feb 04 23:49:59 2008 +1100" }, "message": "virtio: configuration change callback\n\nVarious drivers want to know when their configuration information\nchanges: the balloon driver is the immediate user, but the network\ndriver may one day have a \"carrier\" status as well.\n\nThis introduces that callback (lguest doesn\u0027t use it yet).\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\n" }, { "commit": "18445c4d501b9ab4336f66ef46b092661ddaf336", "tree": "9d23185f207d912942890cf047d1d3200806b401", "parents": [ "a586d4f6016f7139d8c26df0e6927131168d3b5b" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Mon Feb 04 23:49:57 2008 -0500" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Mon Feb 04 23:49:58 2008 +1100" }, "message": "virtio: explicit enable_cb/disable_cb rather than callback return.\n\nIt seems that virtio_net wants to disable callbacks (interrupts) before\ncalling netif_rx_schedule(), so we can\u0027t use the return value to do so.\n\nRename \"restart\" to \"cb_enable\" and introduce \"cb_disable\" hook: callback\nnow returns void, rather than a boolean.\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\n" }, { "commit": "ec3d41c4db4c21164332826ea8d812f94f2f6886", "tree": "9b947e900745cfcc4df7409c6ba5583428964215", "parents": [ "47436aa4ad054c1c7c8231618e86ebd9305308dc" ], "author": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Mon Oct 22 11:03:36 2007 +1000" }, "committer": { "name": "Rusty Russell", "email": "rusty@rustcorp.com.au", "time": "Tue Oct 23 15:49:54 2007 +1000" }, "message": "Virtio interface\n\nThis attempts to implement a \"virtual I/O\" layer which should allow\ncommon drivers to be efficiently used across most virtual I/O\nmechanisms. It will no-doubt need further enhancement.\n\nThe virtio drivers add buffers to virtio queues; as the buffers are consumed\nthe driver \"interrupt\" callbacks are invoked.\n\nThere is also a generic implementation of config space which drivers can query\nto get setup information from the host.\n\nSigned-off-by: Rusty Russell \u003crusty@rustcorp.com.au\u003e\nCc: Dor Laor \u003cdor.laor@qumranet.com\u003e\nCc: Arnd Bergmann \u003carnd@arndb.de\u003e\n" } ] }