linux/fs/exec.c, branch v2.6.34

revert "procfs: provide stack information for threads" and its fixup commits

2010-05-12T00:33:41Z

Originally, commit d899bf7b ("procfs: provide stack information for threads") attempted to introduce a new feature for showing where the threadstack was located and how many pages are being utilized by the stack. Commit c44972f1 ("procfs: disable per-task stack usage on NOMMU") was applied to fix the NO_MMU case. Commit 89240ba0 ("x86, fs: Fix x86 procfs stack information for threads on 64-bit") was applied to fix a bug in ia32 executables being loaded. Commit 9ebd4eba7 ("procfs: fix /proc//stat stack pointer for kernel threads") was applied to fix a bug which had kernel threads printing a userland stack address. Commit 1306d603f ('proc: partially revert "procfs: provide stack information for threads"') was then applied to revert the stack pages being used to solve a significant performance regression. This patch nearly undoes the effect of all these patches. The reason for reverting these is it provides an unusable value in field 28. For x86_64, a fork will result in the task->stack_start value being updated to the current user top of stack and not the stack start address. This unpredictability of the stack_start value makes it worthless. That includes the intended use of showing how much stack space a thread has. Other architectures will get different values. As an example, ia64 gets 0. The do_fork() and copy_process() functions appear to treat the stack_start and stack_size parameters as architecture specific. I only partially reverted c44972f1 ("procfs: disable per-task stack usage on NOMMU") . If I had completely reverted it, I would have had to change mm/Makefile only build pagewalk.o when CONFIG_PROC_PAGE_MONITOR is configured. Since I could not test the builds without significant effort, I decided to not change mm/Makefile. I only partially reverted 89240ba0 ("x86, fs: Fix x86 procfs stack information for threads on 64-bit") . I left the KSTK_ESP() change in place as that seemed worthwhile. Signed-off-by: Robin Holt Cc: Stefani Seibold Cc: KOSAKI Motohiro Cc: Michal Simek Cc: Ingo Molnar Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

coredump: suppress uid comparison test if core output files are pipes

2010-03-06T19:26:46Z

Modify uid check in do_coredump so as to not apply it in the case of pipes. This just got noticed in testing. The end of do_coredump validates the uid of the inode for the created file against the uid of the crashing process to ensure that no one can pre-create a core file with different ownership and grab the information contained in the core when they shouldn' tbe able to. This causes failures when using pipes for a core dumps if the crashing process is not root, which is the uid of the pipe when it is created. The fix is simple. Since the check for matching uid's isn't relevant for pipes (a process can't create a pipe that the uermodehelper code will open anyway), we can just just skip it in the event ispipe is non-zero Reverts a pipe-affecting change which was accidentally made in : commit c46f739dd39db3b07ab5deb4e3ec81e1c04a91af : Author: Ingo Molnar : AuthorDate: Wed Nov 28 13:59:18 2007 +0100 : Commit: Linus Torvalds : CommitDate: Wed Nov 28 10:58:01 2007 -0800 : : vfs: coredumping fix Signed-off-by: Neil Horman Cc: Andi Kleen Cc: Oleg Nesterov Cc: Alan Cox Cc: Al Viro Cc: Ingo Molnar Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

coredump: set ->group_exit_code for other CLONE_VM tasks too

2010-03-06T19:26:46Z

User visible change. do_coredump() kills all threads which share the same ->mm but only the coredumping process gets the proper exit_code. Other tasks which share the same ->mm die "silently" and return status == 0 to parent. This is historical behaviour, not actually a bug. But I think Frank Heckenbach rightly dislikes the current behaviour. Simple test-case: #include #include #include #include int main(void) { int stat; if (!fork()) { if (!vfork()) kill(getpid(), SIGQUIT); } wait(&stat); printf("stat=%x\n", stat); return 0; } Before this patch it prints "stat=0" despite the fact the child was killed by SIGQUIT. After this patch the output is "stat=3" which obviously makes more sense. Even with this patch, only the task which originates the coredumping gets "|= 0x80" if the core was actually dumped, but at least the coredumping signal is visible to do_wait/etc. Reported-by: Frank Heckenbach Signed-off-by: Oleg Nesterov Acked-by: WANG Cong Cc: Roland McGrath Cc: Neil Horman Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

coredump: pass mm->flags as a coredump parameter for consistency

2010-03-06T19:26:46Z

Pass mm->flags as a coredump parameter for consistency. --- 1787 if (mm->core_state || !get_dumpable(mm)) { <- (1) 1788 up_write(&mm->mmap_sem); 1789 put_cred(cred); 1790 goto fail; 1791 } 1792 [...] 1798 if (get_dumpable(mm) == 2) { /* Setuid core dump mode */ <-(2) 1799 flag = O_EXCL; /* Stop rewrite attacks */ 1800 cred->fsuid = 0; /* Dump root private */ 1801 } --- Since dumpable bits are not protected by lock, there is a chance to change these bits between (1) and (2). To solve this issue, this patch copies mm->flags to coredump_params.mm_flags at the beginning of do_coredump() and uses it instead of get_dumpable() while dumping core. This copy is also passed to binfmt->core_dump, since elf*_core_dump() uses dump_filter bits in mm->flags. [akpm@linux-foundation.org: fix merge] Signed-off-by: Masami Hiramatsu Acked-by: Roland McGrath Cc: Hidehiro Kawai Cc: Oleg Nesterov Cc: Ingo Molnar Reviewed-by: KOSAKI Motohiro Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

exec: create initial stack independent of PAGE_SIZE

2010-03-06T19:26:33Z

Currently we create the initial stack based on the PAGE_SIZE. This is unnecessary. This creates this initial stack independent of the PAGE_SIZE. It also bumps up the number of 4k pages allocated from 20 to 32, to align with 64K page systems. Signed-off-by: Michael Neuling Cc: Helge Deller Reviewed-by: KOSAKI Motohiro Cc: Americo Wang Cc: Anton Blanchard Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

fs: use rlimit helpers

2010-03-06T19:26:29Z

Make sure compiler won't do weird things with limits. E.g. fetching them twice may return 2 different values after writable limits are implemented. I.e. either use rlimit helpers added in commit 3e10e716abf3 ("resource: add helpers for fetching rlimits") or ACCESS_ONCE if not applicable. Signed-off-by: Jiri Slaby Cc: Alexander Viro Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm: change anon_vma linking to fix multi-process server scalability issue

2010-03-06T19:26:26Z

The old anon_vma code can lead to scalability issues with heavily forking workloads. Specifically, each anon_vma will be shared between the parent process and all its child processes. In a workload with 1000 child processes and a VMA with 1000 anonymous pages per process that get COWed, this leads to a system with a million anonymous pages in the same anon_vma, each of which is mapped in just one of the 1000 processes. However, the current rmap code needs to walk them all, leading to O(N) scanning complexity for each page. This can result in systems where one CPU is walking the page tables of 1000 processes in page_referenced_one, while all other CPUs are stuck on the anon_vma lock. This leads to catastrophic failure for a benchmark like AIM7, where the total number of processes can reach in the tens of thousands. Real workloads are still a factor 10 less process intensive than AIM7, but they are catching up. This patch changes the way anon_vmas and VMAs are linked, which allows us to associate multiple anon_vmas with a VMA. At fork time, each child process gets its own anon_vmas, in which its COWed pages will be instantiated. The parents' anon_vma is also linked to the VMA, because non-COWed pages could be present in any of the children. This reduces rmap scanning complexity to O(1) for the pages of the 1000 child processes, with O(N) complexity for at most 1/N pages in the system. This reduces the average scanning cost in heavily forking workloads from O(N) to 2. The only real complexity in this patch stems from the fact that linking a VMA to anon_vmas now involves memory allocations. This means vma_adjust can fail, if it needs to attach a VMA to anon_vma structures. This in turn means error handling needs to be added to the calling functions. A second source of complexity is that, because there can be multiple anon_vmas, the anon_vma linking in vma_adjust can no longer be done under "the" anon_vma lock. To prevent the rmap code from walking up an incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h to make sure it is impossible to compile a kernel that needs both symbolic values for the same bitflag. Some test results: Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test box with 16GB RAM and not quite enough IO), the system ends up running >99% in system time, with every CPU on the same anon_vma lock in the pageout code. With these changes, AIM7 hits the cross-over point around 29.7k users. This happens with ~99% IO wait time, there never seems to be any spike in system time. The anon_vma lock contention appears to be resolved. [akpm@linux-foundation.org: cleanups] Signed-off-by: Rik van Riel Cc: KOSAKI Motohiro Cc: Larry Woodman Cc: Lee Schermerhorn Cc: Minchan Kim Cc: Andrea Arcangeli Cc: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

mm: avoid false sharing of mm_counter

2010-03-06T19:26:24Z

Considering the nature of per mm stats, it's the shared object among threads and can be a cache-miss point in the page fault path. This patch adds per-thread cache for mm_counter. RSS value will be counted into a struct in task_struct and synchronized with mm's one at events. Now, in this patch, the event is the number of calls to handle_mm_fault. Per-thread value is added to mm at each 64 calls. rough estimation with small benchmark on parallel thread (2threads) shows [before] 4.5 cache-miss/faults [after] 4.0 cache-miss/faults Anyway, the most contended object is mmap_sem if the number of threads grows. [akpm@linux-foundation.org: coding-style fixes] Signed-off-by: KAMEZAWA Hiroyuki Cc: Minchan Kim Cc: Christoph Lameter Cc: Lee Schermerhorn Cc: David Rientjes Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

fs/exec.c: fix initial stack reservation

2010-02-23T03:50:34Z

803bf5ec259941936262d10ecc84511b76a20921 ("fs/exec.c: restrict initial stack space expansion to rlimit") attempts to limit the initial stack to 20*PAGE_SIZE. Unfortunately, in attempting ensure the stack is not reduced in size, we ended up not changing the stack at all. This size reduction check is not necessary as the expand_stack call does this already. This caused a regression in UML resulting in most guest processes being killed. Signed-off-by: Michael Neuling Reviewed-by: KOSAKI Motohiro Acked-by: WANG Cong Cc: Anton Blanchard Cc: Oleg Nesterov Cc: James Morris Cc: Serge Hallyn Cc: Benjamin Herrenschmidt Cc: Jouni Malinen Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

fs/exec.c: restrict initial stack space expansion to rlimit

2010-02-11T21:59:43Z

When reserving stack space for a new process, make sure we're not attempting to expand the stack by more than rlimit allows. This fixes a bug caused by b6a2fea39318e43fee84fa7b0b90d68bed92d2ba ("mm: variable length argument support") and unmasked by fc63cf237078c86214abcb2ee9926d8ad289da9b ("exec: setup_arg_pages() fails to return errors"). This bug means that when limiting the stack to less the 20*PAGE_SIZE (eg. 80K on 4K pages or 'ulimit -s 79') all processes will be killed before they start. This is particularly bad with 64K pages, where a ulimit below 1280K will kill every process. To test, do: 'ulimit -s 15; ls' before and after the patch is applied. Before it's applied, 'ls' should be killed. After the patch is applied, 'ls' should no longer be killed. A stack limit of 15KB since it's small enough to trigger 20*PAGE_SIZE. Also 15KB not a multiple of PAGE_SIZE, which is a trickier case to handle correctly with this code. 4K pages should be fine to test with. [kosaki.motohiro@jp.fujitsu.com: cleanup] [akpm@linux-foundation.org: cleanup cleanup] Signed-off-by: Michael Neuling Signed-off-by: KOSAKI Motohiro Cc: Americo Wang Cc: Anton Blanchard Cc: Oleg Nesterov Cc: James Morris Cc: Ingo Molnar Cc: Serge Hallyn Cc: Benjamin Herrenschmidt Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds