Change the perf RLIMIT_MEMLOCK accounting to use VM_PINNED. Because the way VM_PINNED works (it hard assumes the entire vma length is accounted) we have to slightly change semantics.
We used to only add to the RLIMIT_MEMLOCK accounting once we were over the per-user limit, now we'll directly account to both. XXX: anon_inode_inode->i_mapping doesn't have AS_UNEVICTABLE set, should it? Cc: Andrew Morton <[email protected]> Cc: Hugh Dickins <[email protected]> Cc: Mel Gorman <[email protected]> Cc: Roland Dreier <[email protected]> Cc: Christoph Lameter <[email protected]> Cc: Thomas Gleixner <[email protected]> Signed-off-by: Peter Zijlstra <[email protected]> --- kernel/events/core.c | 36 ++++++++++++++++-------------------- 1 file changed, 16 insertions(+), 20 deletions(-) --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -4059,13 +4059,12 @@ static const struct vm_operations_struct static int perf_mmap(struct file *file, struct vm_area_struct *vma) { struct perf_event *event = file->private_data; + unsigned long locked, lock_limit, lock_extra; unsigned long user_locked, user_lock_limit; struct user_struct *user = current_user(); - unsigned long locked, lock_limit; - struct ring_buffer *rb; unsigned long vma_size; unsigned long nr_pages; - long user_extra, extra; + struct ring_buffer *rb; int ret = 0, flags = 0; /* @@ -4117,26 +4116,22 @@ static int perf_mmap(struct file *file, goto unlock; } - user_extra = nr_pages + 1; - user_lock_limit = sysctl_perf_event_mlock >> (PAGE_SHIFT - 10); + lock_extra = nr_pages + 1; /* * Increase the limit linearly with more CPUs: */ + user_lock_limit = sysctl_perf_event_mlock >> (PAGE_SHIFT - 10); user_lock_limit *= num_online_cpus(); - user_locked = atomic_long_read(&user->locked_vm) + user_extra; - - extra = 0; - if (user_locked > user_lock_limit) - extra = user_locked - user_lock_limit; + user_locked = atomic_long_read(&user->locked_vm) + lock_extra; lock_limit = rlimit(RLIMIT_MEMLOCK); lock_limit >>= PAGE_SHIFT; - locked = vma->vm_mm->pinned_vm + extra; + locked = mm_locked_pages(vma->vm_mm) + lock_extra; - if ((locked > lock_limit) && perf_paranoid_tracepoint_raw() && - !capable(CAP_IPC_LOCK)) { + if ((user_locked > user_lock_limit && locked > lock_limit) && + perf_paranoid_tracepoint_raw() && !capable(CAP_IPC_LOCK)) { ret = -EPERM; goto unlock; } @@ -4146,7 +4141,7 @@ static int perf_mmap(struct file *file, if (vma->vm_flags & VM_WRITE) flags |= RING_BUFFER_WRITABLE; - rb = rb_alloc(nr_pages, + rb = rb_alloc(nr_pages, event->attr.watermark ? event->attr.wakeup_watermark : 0, event->cpu, flags); @@ -4156,11 +4151,9 @@ static int perf_mmap(struct file *file, } atomic_set(&rb->mmap_count, 1); - rb->mmap_locked = extra; rb->mmap_user = get_current_user(); - atomic_long_add(user_extra, &user->locked_vm); - vma->vm_mm->pinned_vm += extra; + atomic_long_add(lock_extra, &user->locked_vm); ring_buffer_attach(event, rb); @@ -4173,10 +4166,13 @@ static int perf_mmap(struct file *file, mutex_unlock(&event->mmap_mutex); /* - * Since pinned accounting is per vm we cannot allow fork() to copy our - * vma. + * VM_PINNED - this memory is pinned as we need to write to it from + * pretty much any context and cannot page. + * VM_DONTCOPY - don't share over fork() + * VM_DONTEXPAND - its not stack + * VM_DONTDUMP - ... */ - vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND | VM_DONTDUMP; + vma->vm_flags |= VM_PINNED | VM_DONTCOPY | VM_DONTEXPAND | VM_DONTDUMP; vma->vm_ops = &perf_mmap_vmops; return ret; -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
