[PATCH] uml: TLB operation batching
This adds VM op batching to skas0. Rather than having a context switch to and
from the userspace stub for each address space change, we write a number of
operations to the stub data page and invoke a different stub which loops over
them and executes them all in one go.
The operations are stored as [ system call number, arg1, arg2, ... ] tuples.
The set is terminated by a system call number of 0. Single operations, i.e.
page faults, are handled in the old way, since that is slightly more
efficient.
For a kernel build, a minority (~1/4) of the operations are part of a set.
These sets averaged ~100 in length, so for this quarter, the context switching
overhead is greatly reduced.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Cc: Paolo Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
diff --git a/arch/um/kernel/skas/include/skas.h b/arch/um/kernel/skas/include/skas.h
index d983ea8..e91064b 100644
--- a/arch/um/kernel/skas/include/skas.h
+++ b/arch/um/kernel/skas/include/skas.h
@@ -24,11 +24,14 @@
extern void remove_sigstack(void);
extern void new_thread_handler(int sig);
extern void handle_syscall(union uml_pt_regs *regs);
-extern int map(struct mm_id * mm_idp, unsigned long virt, unsigned long len,
- int r, int w, int x, int phys_fd, unsigned long long offset);
-extern int unmap(struct mm_id * mm_idp, void *addr, unsigned long len);
-extern int protect(struct mm_id * mm_idp, unsigned long addr,
- unsigned long len, int r, int w, int x);
+extern void *map(struct mm_id * mm_idp, unsigned long virt,
+ unsigned long len, int r, int w, int x, int phys_fd,
+ unsigned long long offset, int done, void *data);
+extern void *unmap(struct mm_id * mm_idp, void *addr,
+ unsigned long len, int done, void *data);
+extern void *protect(struct mm_id * mm_idp, unsigned long addr,
+ unsigned long len, int r, int w, int x, int done,
+ void *data);
extern void user_signal(int sig, union uml_pt_regs *regs, int pid);
extern int new_mm(int from);
extern int start_userspace(unsigned long stub_stack);
diff --git a/arch/um/kernel/skas/mem_user.c b/arch/um/kernel/skas/mem_user.c
index b0980ff..c976320 100644
--- a/arch/um/kernel/skas/mem_user.c
+++ b/arch/um/kernel/skas/mem_user.c
@@ -25,12 +25,14 @@
#include "sysdep/stub.h"
#include "skas.h"
-extern unsigned long syscall_stub, __syscall_stub_start;
+extern unsigned long syscall_stub, batch_syscall_stub, __syscall_stub_start;
extern void wait_stub_done(int pid, int sig, char * fname);
-static long run_syscall_stub(struct mm_id * mm_idp, int syscall,
- unsigned long *args)
+int single_count = 0;
+
+static long one_syscall_stub(struct mm_id * mm_idp, int syscall,
+ unsigned long *args)
{
int n, pid = mm_idp->u.pid;
unsigned long regs[MAX_REG_NR];
@@ -49,18 +51,80 @@
regs[REGS_SYSCALL_ARG6] = args[5];
n = ptrace_setregs(pid, regs);
if(n < 0){
- printk("run_syscall_stub : PTRACE_SETREGS failed, "
+ printk("one_syscall_stub : PTRACE_SETREGS failed, "
+ "errno = %d\n", n);
+ return(n);
+ }
+
+ wait_stub_done(pid, 0, "one_syscall_stub");
+
+ return(*((unsigned long *) mm_idp->stack));
+}
+
+int multi_count = 0;
+int multi_op_count = 0;
+
+static long many_syscall_stub(struct mm_id * mm_idp, int syscall,
+ unsigned long *args, int done, void **addr_out)
+{
+ unsigned long regs[MAX_REG_NR], *stack;
+ int n, pid = mm_idp->u.pid;
+
+ stack = *addr_out;
+ if(stack == NULL)
+ stack = (unsigned long *) current_stub_stack();
+ *stack++ = syscall;
+ *stack++ = args[0];
+ *stack++ = args[1];
+ *stack++ = args[2];
+ *stack++ = args[3];
+ *stack++ = args[4];
+ *stack++ = args[5];
+ *stack = 0;
+ multi_op_count++;
+
+ if(!done && ((((unsigned long) stack) & ~PAGE_MASK) <
+ PAGE_SIZE - 8 * sizeof(long))){
+ *addr_out = stack;
+ return 0;
+ }
+
+ multi_count++;
+ get_safe_registers(regs);
+ regs[REGS_IP_INDEX] = UML_CONFIG_STUB_CODE +
+ ((unsigned long) &batch_syscall_stub -
+ (unsigned long) &__syscall_stub_start);
+ regs[REGS_SP_INDEX] = UML_CONFIG_STUB_DATA;
+
+ n = ptrace_setregs(pid, regs);
+ if(n < 0){
+ printk("many_syscall_stub : PTRACE_SETREGS failed, "
"errno = %d\n", n);
return(n);
}
- wait_stub_done(pid, 0, "run_syscall_stub");
+ wait_stub_done(pid, 0, "many_syscall_stub");
+ stack = (unsigned long *) mm_idp->stack;
- return(*((unsigned long *) mm_idp->stack));
+ *addr_out = stack;
+ return(*stack);
}
-int map(struct mm_id *mm_idp, unsigned long virt, unsigned long len,
- int r, int w, int x, int phys_fd, unsigned long long offset)
+static long run_syscall_stub(struct mm_id * mm_idp, int syscall,
+ unsigned long *args, void **addr, int done)
+{
+ long res;
+
+ if((*addr == NULL) && done)
+ res = one_syscall_stub(mm_idp, syscall, args);
+ else res = many_syscall_stub(mm_idp, syscall, args, done, addr);
+
+ return res;
+}
+
+void *map(struct mm_id * mm_idp, unsigned long virt, unsigned long len,
+ int r, int w, int x, int phys_fd, unsigned long long offset,
+ int done, void *data)
{
int prot, n;
@@ -70,6 +134,7 @@
if(proc_mm){
struct proc_mm_op map;
int fd = mm_idp->u.mm_fd;
+
map = ((struct proc_mm_op) { .op = MM_MMAP,
.u =
{ .mmap =
@@ -91,21 +156,24 @@
MAP_SHARED | MAP_FIXED, phys_fd,
MMAP_OFFSET(offset) };
- res = run_syscall_stub(mm_idp, STUB_MMAP_NR, args);
+ res = run_syscall_stub(mm_idp, STUB_MMAP_NR, args,
+ &data, done);
if((void *) res == MAP_FAILED)
printk("mmap stub failed, errno = %d\n", res);
}
- return 0;
+ return data;
}
-int unmap(struct mm_id *mm_idp, void *addr, unsigned long len)
+void *unmap(struct mm_id * mm_idp, void *addr, unsigned long len, int done,
+ void *data)
{
int n;
if(proc_mm){
struct proc_mm_op unmap;
int fd = mm_idp->u.mm_fd;
+
unmap = ((struct proc_mm_op) { .op = MM_MUNMAP,
.u =
{ .munmap =
@@ -113,28 +181,25 @@
(unsigned long) addr,
.len = len } } } );
n = os_write_file(fd, &unmap, sizeof(unmap));
- if(n != sizeof(unmap)) {
- if(n < 0)
- return(n);
- else if(n > 0)
- return(-EIO);
- }
+ if(n != sizeof(unmap))
+ printk("unmap - proc_mm write returned %d\n", n);
}
else {
int res;
unsigned long args[] = { (unsigned long) addr, len, 0, 0, 0,
0 };
- res = run_syscall_stub(mm_idp, __NR_munmap, args);
+ res = run_syscall_stub(mm_idp, __NR_munmap, args,
+ &data, done);
if(res < 0)
printk("munmap stub failed, errno = %d\n", res);
}
- return(0);
+ return data;
}
-int protect(struct mm_id *mm_idp, unsigned long addr, unsigned long len,
- int r, int w, int x)
+void *protect(struct mm_id * mm_idp, unsigned long addr, unsigned long len,
+ int r, int w, int x, int done, void *data)
{
struct proc_mm_op protect;
int prot, n;
@@ -160,12 +225,13 @@
int res;
unsigned long args[] = { addr, len, prot, 0, 0, 0 };
- res = run_syscall_stub(mm_idp, __NR_mprotect, args);
+ res = run_syscall_stub(mm_idp, __NR_mprotect, args,
+ &data, done);
if(res < 0)
panic("mprotect stub failed, errno = %d\n", res);
}
- return(0);
+ return data;
}
void before_mem_skas(unsigned long unused)
diff --git a/arch/um/kernel/skas/tlb.c b/arch/um/kernel/skas/tlb.c
index 6230999..4b5fd20 100644
--- a/arch/um/kernel/skas/tlb.c
+++ b/arch/um/kernel/skas/tlb.c
@@ -18,7 +18,8 @@
#include "os.h"
#include "tlb.h"
-static void do_ops(union mm_context *mmu, struct host_vm_op *ops, int last)
+static void *do_ops(union mm_context *mmu, struct host_vm_op *ops, int last,
+ int finished, void *flush)
{
struct host_vm_op *op;
int i;
@@ -27,24 +28,28 @@
op = &ops[i];
switch(op->type){
case MMAP:
- map(&mmu->skas.id, op->u.mmap.addr, op->u.mmap.len,
- op->u.mmap.r, op->u.mmap.w, op->u.mmap.x,
- op->u.mmap.fd, op->u.mmap.offset);
+ flush = map(&mmu->skas.id, op->u.mmap.addr,
+ op->u.mmap.len, op->u.mmap.r, op->u.mmap.w,
+ op->u.mmap.x, op->u.mmap.fd,
+ op->u.mmap.offset, finished, flush);
break;
case MUNMAP:
- unmap(&mmu->skas.id, (void *) op->u.munmap.addr,
- op->u.munmap.len);
+ flush = unmap(&mmu->skas.id, (void *) op->u.munmap.addr,
+ op->u.munmap.len, finished, flush);
break;
case MPROTECT:
- protect(&mmu->skas.id, op->u.mprotect.addr,
- op->u.mprotect.len, op->u.mprotect.r,
- op->u.mprotect.w, op->u.mprotect.x);
+ flush = protect(&mmu->skas.id, op->u.mprotect.addr,
+ op->u.mprotect.len, op->u.mprotect.r,
+ op->u.mprotect.w, op->u.mprotect.x,
+ finished, flush);
break;
default:
printk("Unknown op type %d in do_ops\n", op->type);
break;
}
}
+
+ return flush;
}
extern int proc_mm;