Re: [PATCH -mmotm 3/3] memcg: dirty pages instrumentation

[Kirill A. Shutemov]

On Mon, Mar 1, 2010 at 11:23 PM, Andrea Righi <arighi@xxxxxxxxxxx> wrote:
> Apply the cgroup dirty pages accounting and limiting infrastructure to
> the opportune kernel functions.
>
> Signed-off-by: Andrea Righi <arighi@xxxxxxxxxxx>
> ---
> Âfs/fuse/file.c   Â|  Â5 +++
> Âfs/nfs/write.c   Â|  Â4 ++
> Âfs/nilfs2/segment.c | Â 10 +++++-
> Âmm/filemap.c    Â|  Â1 +
> Âmm/page-writeback.c | Â 84 ++++++++++++++++++++++++++++++++------------------
> Âmm/rmap.c      |  Â4 +-
> Âmm/truncate.c    |  Â2 +
> Â7 files changed, 76 insertions(+), 34 deletions(-)
>
> diff --git a/fs/fuse/file.c b/fs/fuse/file.c
> index a9f5e13..dbbdd53 100644
> --- a/fs/fuse/file.c
> +++ b/fs/fuse/file.c
> @@ -11,6 +11,7 @@
> Â#include <linux/pagemap.h>
> Â#include <linux/slab.h>
> Â#include <linux/kernel.h>
> +#include <linux/memcontrol.h>
> Â#include <linux/sched.h>
> Â#include <linux/module.h>
>
> @@ -1129,6 +1130,8 @@ static void fuse_writepage_finish(struct fuse_conn *fc, struct fuse_req *req)
>
> Â Â Â Âlist_del(&req->writepages_entry);
> Â Â Â Âdec_bdi_stat(bdi, BDI_WRITEBACK);
> + Â Â Â mem_cgroup_update_stat(req->pages[0],
> + Â Â Â Â Â Â Â Â Â Â Â MEM_CGROUP_STAT_WRITEBACK_TEMP, -1);
> Â Â Â Âdec_zone_page_state(req->pages[0], NR_WRITEBACK_TEMP);
> Â Â Â Âbdi_writeout_inc(bdi);
> Â Â Â Âwake_up(&fi->page_waitq);
> @@ -1240,6 +1243,8 @@ static int fuse_writepage_locked(struct page *page)
> Â Â Â Âreq->inode = inode;
>
> Â Â Â Âinc_bdi_stat(mapping->backing_dev_info, BDI_WRITEBACK);
> + Â Â Â mem_cgroup_update_stat(tmp_page,
> + Â Â Â Â Â Â Â Â Â Â Â MEM_CGROUP_STAT_WRITEBACK_TEMP, 1);
> Â Â Â Âinc_zone_page_state(tmp_page, NR_WRITEBACK_TEMP);
> Â Â Â Âend_page_writeback(page);
>
> diff --git a/fs/nfs/write.c b/fs/nfs/write.c
> index b753242..7316f7a 100644
> --- a/fs/nfs/write.c
> +++ b/fs/nfs/write.c
> @@ -439,6 +439,7 @@ nfs_mark_request_commit(struct nfs_page *req)
> Â Â Â Â Â Â Â Â Â Â Â Âreq->wb_index,
> Â Â Â Â Â Â Â Â Â Â Â ÂNFS_PAGE_TAG_COMMIT);
> Â Â Â Âspin_unlock(&inode->i_lock);
> + Â Â Â mem_cgroup_update_stat(req->wb_page, MEM_CGROUP_STAT_UNSTABLE_NFS, 1);
> Â Â Â Âinc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
> Â Â Â Âinc_bdi_stat(req->wb_page->mapping->backing_dev_info, BDI_UNSTABLE);
> Â Â Â Â__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
> @@ -450,6 +451,7 @@ nfs_clear_request_commit(struct nfs_page *req)
> Â Â Â Âstruct page *page = req->wb_page;
>
> Â Â Â Âif (test_and_clear_bit(PG_CLEAN, &(req)->wb_flags)) {
> + Â Â Â Â Â Â Â mem_cgroup_update_stat(page, MEM_CGROUP_STAT_UNSTABLE_NFS, -1);
> Â Â Â Â Â Â Â Âdec_zone_page_state(page, NR_UNSTABLE_NFS);
> Â Â Â Â Â Â Â Âdec_bdi_stat(page->mapping->backing_dev_info, BDI_UNSTABLE);
> Â Â Â Â Â Â Â Âreturn 1;
> @@ -1273,6 +1275,8 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how)
> Â Â Â Â Â Â Â Âreq = nfs_list_entry(head->next);
> Â Â Â Â Â Â Â Ânfs_list_remove_request(req);
> Â Â Â Â Â Â Â Ânfs_mark_request_commit(req);
> + Â Â Â Â Â Â Â mem_cgroup_update_stat(req->wb_page,
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â MEM_CGROUP_STAT_UNSTABLE_NFS, -1);
> Â Â Â Â Â Â Â Âdec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
> Â Â Â Â Â Â Â Âdec_bdi_stat(req->wb_page->mapping->backing_dev_info,
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â ÂBDI_UNSTABLE);
> diff --git a/fs/nilfs2/segment.c b/fs/nilfs2/segment.c
> index ada2f1b..aef6d13 100644
> --- a/fs/nilfs2/segment.c
> +++ b/fs/nilfs2/segment.c
> @@ -1660,8 +1660,11 @@ nilfs_copy_replace_page_buffers(struct page *page, struct list_head *out)
> Â Â Â Â} while (bh = bh->b_this_page, bh2 = bh2->b_this_page, bh != head);
> Â Â Â Âkunmap_atomic(kaddr, KM_USER0);
>
> - Â Â Â if (!TestSetPageWriteback(clone_page))
> + Â Â Â if (!TestSetPageWriteback(clone_page)) {
> + Â Â Â Â Â Â Â mem_cgroup_update_stat(clone_page,

s/clone_page/page/

And #include <linux/memcontrol.h> is missed.

> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â MEM_CGROUP_STAT_WRITEBACK, 1);
> Â Â Â Â Â Â Â Âinc_zone_page_state(clone_page, NR_WRITEBACK);
> + Â Â Â }
> Â Â Â Âunlock_page(clone_page);
>
> Â Â Â Âreturn 0;
> @@ -1783,8 +1786,11 @@ static void __nilfs_end_page_io(struct page *page, int err)
> Â Â Â Â}
>
> Â Â Â Âif (buffer_nilfs_allocated(page_buffers(page))) {
> - Â Â Â Â Â Â Â if (TestClearPageWriteback(page))
> + Â Â Â Â Â Â Â if (TestClearPageWriteback(page)) {
> + Â Â Â Â Â Â Â Â Â Â Â mem_cgroup_update_stat(clone_page,
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â MEM_CGROUP_STAT_WRITEBACK, -1);
> Â Â Â Â Â Â Â Â Â Â Â Âdec_zone_page_state(page, NR_WRITEBACK);
> + Â Â Â Â Â Â Â }
> Â Â Â Â} else
> Â Â Â Â Â Â Â Âend_page_writeback(page);
> Â}
> diff --git a/mm/filemap.c b/mm/filemap.c
> index fe09e51..f85acae 100644
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -135,6 +135,7 @@ void __remove_from_page_cache(struct page *page)
> Â Â Â Â * having removed the page entirely.
> Â Â Â Â */
> Â Â Â Âif (PageDirty(page) && mapping_cap_account_dirty(mapping)) {
> + Â Â Â Â Â Â Â mem_cgroup_update_stat(page, MEM_CGROUP_STAT_FILE_DIRTY, -1);
> Â Â Â Â Â Â Â Âdec_zone_page_state(page, NR_FILE_DIRTY);
> Â Â Â Â Â Â Â Âdec_bdi_stat(mapping->backing_dev_info, BDI_DIRTY);
> Â Â Â Â}
> diff --git a/mm/page-writeback.c b/mm/page-writeback.c
> index 5a0f8f3..d83f41c 100644
> --- a/mm/page-writeback.c
> +++ b/mm/page-writeback.c
> @@ -137,13 +137,14 @@ static struct prop_descriptor vm_dirties;
> Â*/
> Âstatic int calc_period_shift(void)
> Â{
> - Â Â Â unsigned long dirty_total;
> + Â Â Â unsigned long dirty_total, dirty_bytes;
>
> - Â Â Â if (vm_dirty_bytes)
> - Â Â Â Â Â Â Â dirty_total = vm_dirty_bytes / PAGE_SIZE;
> + Â Â Â dirty_bytes = mem_cgroup_dirty_bytes();
> + Â Â Â if (dirty_bytes)
> + Â Â Â Â Â Â Â dirty_total = dirty_bytes / PAGE_SIZE;
> Â Â Â Âelse
> - Â Â Â Â Â Â Â dirty_total = (vm_dirty_ratio * determine_dirtyable_memory()) /
> - Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â 100;
> + Â Â Â Â Â Â Â dirty_total = (mem_cgroup_dirty_ratio() *
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â determine_dirtyable_memory()) / 100;
> Â Â Â Âreturn 2 + ilog2(dirty_total - 1);
> Â}
>
> @@ -408,14 +409,16 @@ static unsigned long highmem_dirtyable_memory(unsigned long total)
> Â*/
> Âunsigned long determine_dirtyable_memory(void)
> Â{
> - Â Â Â unsigned long x;
> -
> - Â Â Â x = global_page_state(NR_FREE_PAGES) + global_reclaimable_pages();
> + Â Â Â unsigned long memory;
> + Â Â Â s64 memcg_memory;
>
> + Â Â Â memory = global_page_state(NR_FREE_PAGES) + global_reclaimable_pages();
> Â Â Â Âif (!vm_highmem_is_dirtyable)
> - Â Â Â Â Â Â Â x -= highmem_dirtyable_memory(x);
> -
> - Â Â Â return x + 1; Â /* Ensure that we never return 0 */
> + Â Â Â Â Â Â Â memory -= highmem_dirtyable_memory(memory);
> + Â Â Â memcg_memory = mem_cgroup_page_stat(MEMCG_NR_DIRTYABLE_PAGES);
> + Â Â Â if (memcg_memory < 0)
> + Â Â Â Â Â Â Â return memory + 1;
> + Â Â Â return min((unsigned long)memcg_memory, memory + 1);
> Â}
>
> Âvoid
> @@ -423,26 +426,28 @@ get_dirty_limits(unsigned long *pbackground, unsigned long *pdirty,
> Â Â Â Â Â Â Â Â unsigned long *pbdi_dirty, struct backing_dev_info *bdi)
> Â{
> Â Â Â Âunsigned long background;
> - Â Â Â unsigned long dirty;
> + Â Â Â unsigned long dirty, dirty_bytes, dirty_background;
> Â Â Â Âunsigned long available_memory = determine_dirtyable_memory();
> Â Â Â Âstruct task_struct *tsk;
>
> - Â Â Â if (vm_dirty_bytes)
> - Â Â Â Â Â Â Â dirty = DIV_ROUND_UP(vm_dirty_bytes, PAGE_SIZE);
> + Â Â Â dirty_bytes = mem_cgroup_dirty_bytes();
> + Â Â Â if (dirty_bytes)
> + Â Â Â Â Â Â Â dirty = DIV_ROUND_UP(dirty_bytes, PAGE_SIZE);
> Â Â Â Âelse {
> Â Â Â Â Â Â Â Âint dirty_ratio;
>
> - Â Â Â Â Â Â Â dirty_ratio = vm_dirty_ratio;
> + Â Â Â Â Â Â Â dirty_ratio = mem_cgroup_dirty_ratio();
> Â Â Â Â Â Â Â Âif (dirty_ratio < 5)
> Â Â Â Â Â Â Â Â Â Â Â Âdirty_ratio = 5;
> Â Â Â Â Â Â Â Âdirty = (dirty_ratio * available_memory) / 100;
> Â Â Â Â}
>
> - Â Â Â if (dirty_background_bytes)
> - Â Â Â Â Â Â Â background = DIV_ROUND_UP(dirty_background_bytes, PAGE_SIZE);
> + Â Â Â dirty_background = mem_cgroup_dirty_background_bytes();
> + Â Â Â if (dirty_background)
> + Â Â Â Â Â Â Â background = DIV_ROUND_UP(dirty_background, PAGE_SIZE);
> Â Â Â Âelse
> - Â Â Â Â Â Â Â background = (dirty_background_ratio * available_memory) / 100;
> -
> + Â Â Â Â Â Â Â background = (mem_cgroup_dirty_background_ratio() *
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â available_memory) / 100;
> Â Â Â Âif (background >= dirty)
> Â Â Â Â Â Â Â Âbackground = dirty / 2;
> Â Â Â Âtsk = current;
> @@ -508,9 +513,13 @@ static void balance_dirty_pages(struct address_space *mapping,
> Â Â Â Â Â Â Â Âget_dirty_limits(&background_thresh, &dirty_thresh,
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â&bdi_thresh, bdi);
>
> - Â Â Â Â Â Â Â nr_reclaimable = global_page_state(NR_FILE_DIRTY) +
> + Â Â Â Â Â Â Â nr_reclaimable = mem_cgroup_page_stat(MEMCG_NR_RECLAIM_PAGES);
> + Â Â Â Â Â Â Â nr_writeback = mem_cgroup_page_stat(MEMCG_NR_WRITEBACK);
> + Â Â Â Â Â Â Â if ((nr_reclaimable < 0) || (nr_writeback < 0)) {
> + Â Â Â Â Â Â Â Â Â Â Â nr_reclaimable = global_page_state(NR_FILE_DIRTY) +
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Âglobal_page_state(NR_UNSTABLE_NFS);
> - Â Â Â Â Â Â Â nr_writeback = global_page_state(NR_WRITEBACK);
> + Â Â Â Â Â Â Â Â Â Â Â nr_writeback = global_page_state(NR_WRITEBACK);
> + Â Â Â Â Â Â Â }
>
> Â Â Â Â Â Â Â Âbdi_nr_reclaimable = bdi_stat(bdi, BDI_DIRTY);
> Â Â Â Â Â Â Â Âif (bdi_cap_account_unstable(bdi)) {
> @@ -611,10 +620,12 @@ static void balance_dirty_pages(struct address_space *mapping,
> Â Â Â Â * In normal mode, we start background writeout at the lower
> Â Â Â Â * background_thresh, to keep the amount of dirty memory low.
> Â Â Â Â */
> + Â Â Â nr_reclaimable = mem_cgroup_page_stat(MEMCG_NR_RECLAIM_PAGES);
> + Â Â Â if (nr_reclaimable < 0)
> + Â Â Â Â Â Â Â nr_reclaimable = global_page_state(NR_FILE_DIRTY) +
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â global_page_state(NR_UNSTABLE_NFS);
> Â Â Â Âif ((laptop_mode && pages_written) ||
> - Â Â Â Â Â (!laptop_mode && ((global_page_state(NR_FILE_DIRTY)
> - Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â+ global_page_state(NR_UNSTABLE_NFS))
> - Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â > background_thresh)))
> + Â Â Â Â Â (!laptop_mode && (nr_reclaimable > background_thresh)))
> Â Â Â Â Â Â Â Âbdi_start_writeback(bdi, NULL, 0);
> Â}
>
> @@ -678,6 +689,8 @@ void throttle_vm_writeout(gfp_t gfp_mask)
> Â Â Â Âunsigned long dirty_thresh;
>
> Â Â Â Â for ( ; ; ) {
> + Â Â Â Â Â Â Â unsigned long dirty;
> +
> Â Â Â Â Â Â Â Âget_dirty_limits(&background_thresh, &dirty_thresh, NULL, NULL);
>
> Â Â Â Â Â Â Â Â /*
> @@ -686,10 +699,14 @@ void throttle_vm_writeout(gfp_t gfp_mask)
> Â Â Â Â Â Â Â Â Â*/
> Â Â Â Â Â Â Â Â dirty_thresh += dirty_thresh / 10; Â Â Â/* wheeee... */
>
> - Â Â Â Â Â Â Â Âif (global_page_state(NR_UNSTABLE_NFS) +
> - Â Â Â Â Â Â Â Â Â Â Â global_page_state(NR_WRITEBACK) <= dirty_thresh)
> - Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â break;
> - Â Â Â Â Â Â Â Âcongestion_wait(BLK_RW_ASYNC, HZ/10);
> +
> + Â Â Â Â Â Â Â dirty = mem_cgroup_page_stat(MEMCG_NR_DIRTY_WRITEBACK_PAGES);
> + Â Â Â Â Â Â Â if (dirty < 0)
> + Â Â Â Â Â Â Â Â Â Â Â dirty = global_page_state(NR_UNSTABLE_NFS) +
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â global_page_state(NR_WRITEBACK);
> + Â Â Â Â Â Â Â if (dirty <= dirty_thresh)
> + Â Â Â Â Â Â Â Â Â Â Â break;
> + Â Â Â Â Â Â Â congestion_wait(BLK_RW_ASYNC, HZ/10);
>
> Â Â Â Â Â Â Â Â/*
> Â Â Â Â Â Â Â Â * The caller might hold locks which can prevent IO completion
> @@ -1096,6 +1113,7 @@ int __set_page_dirty_no_writeback(struct page *page)
> Âvoid account_page_dirtied(struct page *page, struct address_space *mapping)
> Â{
> Â Â Â Âif (mapping_cap_account_dirty(mapping)) {
> + Â Â Â Â Â Â Â mem_cgroup_update_stat(page, MEM_CGROUP_STAT_FILE_DIRTY, 1);
> Â Â Â Â Â Â Â Â__inc_zone_page_state(page, NR_FILE_DIRTY);
> Â Â Â Â Â Â Â Â__inc_bdi_stat(mapping->backing_dev_info, BDI_DIRTY);
> Â Â Â Â Â Â Â Âtask_dirty_inc(current);
> @@ -1297,6 +1315,8 @@ int clear_page_dirty_for_io(struct page *page)
> Â Â Â Â Â Â Â Â * for more comments.
> Â Â Â Â Â Â Â Â */
> Â Â Â Â Â Â Â Âif (TestClearPageDirty(page)) {
> + Â Â Â Â Â Â Â Â Â Â Â mem_cgroup_update_stat(page,
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â MEM_CGROUP_STAT_FILE_DIRTY, -1);
> Â Â Â Â Â Â Â Â Â Â Â Âdec_zone_page_state(page, NR_FILE_DIRTY);
> Â Â Â Â Â Â Â Â Â Â Â Âdec_bdi_stat(mapping->backing_dev_info,
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â ÂBDI_DIRTY);
> @@ -1332,8 +1352,10 @@ int test_clear_page_writeback(struct page *page)
> Â Â Â Â} else {
> Â Â Â Â Â Â Â Âret = TestClearPageWriteback(page);
> Â Â Â Â}
> - Â Â Â if (ret)
> + Â Â Â if (ret) {
> + Â Â Â Â Â Â Â mem_cgroup_update_stat(page, MEM_CGROUP_STAT_WRITEBACK, -1);
> Â Â Â Â Â Â Â Âdec_zone_page_state(page, NR_WRITEBACK);
> + Â Â Â }
> Â Â Â Âreturn ret;
> Â}
>
> @@ -1363,8 +1385,10 @@ int test_set_page_writeback(struct page *page)
> Â Â Â Â} else {
> Â Â Â Â Â Â Â Âret = TestSetPageWriteback(page);
> Â Â Â Â}
> - Â Â Â if (!ret)
> + Â Â Â if (!ret) {
> + Â Â Â Â Â Â Â mem_cgroup_update_stat(page, MEM_CGROUP_STAT_WRITEBACK, 1);
> Â Â Â Â Â Â Â Âinc_zone_page_state(page, NR_WRITEBACK);
> + Â Â Â }
> Â Â Â Âreturn ret;
>
> Â}
> diff --git a/mm/rmap.c b/mm/rmap.c
> index 4d2fb93..8d74335 100644
> --- a/mm/rmap.c
> +++ b/mm/rmap.c
> @@ -832,7 +832,7 @@ void page_add_file_rmap(struct page *page)
> Â{
> Â Â Â Âif (atomic_inc_and_test(&page->_mapcount)) {
> Â Â Â Â Â Â Â Â__inc_zone_page_state(page, NR_FILE_MAPPED);
> - Â Â Â Â Â Â Â mem_cgroup_update_file_mapped(page, 1);
> + Â Â Â Â Â Â Â mem_cgroup_update_stat(page, MEM_CGROUP_STAT_FILE_MAPPED, 1);
> Â Â Â Â}
> Â}
>
> @@ -864,7 +864,7 @@ void page_remove_rmap(struct page *page)
> Â Â Â Â Â Â Â Â__dec_zone_page_state(page, NR_ANON_PAGES);
> Â Â Â Â} else {
> Â Â Â Â Â Â Â Â__dec_zone_page_state(page, NR_FILE_MAPPED);
> - Â Â Â Â Â Â Â mem_cgroup_update_file_mapped(page, -1);
> + Â Â Â Â Â Â Â mem_cgroup_update_stat(page, MEM_CGROUP_STAT_FILE_MAPPED, -1);
> Â Â Â Â}
> Â Â Â Â/*
> Â Â Â Â * It would be tidy to reset the PageAnon mapping here,
> diff --git a/mm/truncate.c b/mm/truncate.c
> index 2466e0c..5f437e7 100644
> --- a/mm/truncate.c
> +++ b/mm/truncate.c
> @@ -73,6 +73,8 @@ void cancel_dirty_page(struct page *page, unsigned int account_size)
> Â Â Â Âif (TestClearPageDirty(page)) {
> Â Â Â Â Â Â Â Âstruct address_space *mapping = page->mapping;
> Â Â Â Â Â Â Â Âif (mapping && mapping_cap_account_dirty(mapping)) {
> + Â Â Â Â Â Â Â Â Â Â Â mem_cgroup_update_stat(page,
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â MEM_CGROUP_STAT_FILE_DIRTY, -1);
> Â Â Â Â Â Â Â Â Â Â Â Âdec_zone_page_state(page, NR_FILE_DIRTY);
> Â Â Â Â Â Â Â Â Â Â Â Âdec_bdi_stat(mapping->backing_dev_info,
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â ÂBDI_DIRTY);
> --
> 1.6.3.3
>
>
èº{.nÇ+‰·Ÿ®‰­†+%ŠËlzwm…ébëæìr¸›zX§»®w¥Š{ayºÊÚë,j­¢f£¢·hš‹àz¹®w¥¢¸¢·¦j:+v‰¨ŠwèjØm¶Ÿÿ¾«‘êçzZ+ƒùšŽŠÝj"ú!¶iO•æ¬z·švØ^¶m§ÿðÃnÆàþY&—