| Message ID | 20200804161755.10100-2-willy@infradead.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Remove nrexceptional tracking | expand |
On Tue, Aug 04, 2020 at 05:17:52PM +0100, Matthew Wilcox (Oracle) wrote: > diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h > index 484a36185bb5..a474a92a2a72 100644 > --- a/include/linux/pagemap.h > +++ b/include/linux/pagemap.h > @@ -18,6 +18,11 @@ > > struct pagevec; > > +static inline bool page_cache_empty(struct address_space *mapping) > +{ > + return xa_empty(&mapping->i_pages); What about something like bool empty = xa_empty(&mapping->i_pages); VM_BUG_ON(empty && mapping->nrpages); return empty; ? > +} > + > /* > * Bits in mapping->flags. > */
On Fri, Aug 07, 2020 at 02:24:00AM +0300, Kirill A. Shutemov wrote: > On Tue, Aug 04, 2020 at 05:17:52PM +0100, Matthew Wilcox (Oracle) wrote: > > diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h > > index 484a36185bb5..a474a92a2a72 100644 > > --- a/include/linux/pagemap.h > > +++ b/include/linux/pagemap.h > > @@ -18,6 +18,11 @@ > > > > struct pagevec; > > > > +static inline bool page_cache_empty(struct address_space *mapping) > > +{ > > + return xa_empty(&mapping->i_pages); > > What about something like > > bool empty = xa_empty(&mapping->i_pages); > VM_BUG_ON(empty && mapping->nrpages); > return empty; I tried this and it's triggered by generic/418. The problem is that it's called when the pagecache lock isn't held (by invalidate_inode_pages2_range), so it's possible for xa_empty() to return true, then a page be added to the page cache, and mapping->pages be incremented to 1. That seems to be what's happened here: (gdb) p/x *(struct address_space *)0xffff88804b21b360 $2 = {host = 0xffff88804b21b200, i_pages = {xa_lock = {{rlock = {raw_lock = {{ val = {counter = 0x0}, {locked = 0x0, pending = 0x0}, { locked_pending = 0x0, tail = 0x0}}}}}}, xa_flags = 0x21, * xa_head = 0xffffea0001e187c0}, gfp_mask = 0x100c4a, i_mmap_writable = { counter = 0x0}, nr_thps = {counter = 0x0}, i_mmap = {rb_root = { rb_node = 0x0}, rb_leftmost = 0x0}, i_mmap_rwsem = {count = { counter = 0x0}, owner = {counter = 0x0}, osq = {tail = {counter = 0x0}}, wait_lock = {raw_lock = {{val = {counter = 0x0}, {locked = 0x0, pending = 0x0}, {locked_pending = 0x0, tail = 0x0}}}}, wait_list = {next = 0xffff88804b21b3b0, prev = 0xffff88804b21b3b0}}, * nrpages = 0x1, writeback_index = 0x0, a_ops = 0xffffffff81c2ed60, flags = 0x40, wb_err = 0x0, private_lock = {{rlock = {raw_lock = {{val = { counter = 0x0}, {locked = 0x0, pending = 0x0}, { locked_pending = 0x0, tail = 0x0}}}}}}, private_list = { next = 0xffff88804b21b3e8, prev = 0xffff88804b21b3e8}, private_data = 0x0} (marked the critical lines with *)
diff --git a/fs/block_dev.c b/fs/block_dev.c index 0ae656e022fd..2a77bd2c6144 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -79,7 +79,7 @@ static void kill_bdev(struct block_device *bdev) { struct address_space *mapping = bdev->bd_inode->i_mapping; - if (mapping->nrpages == 0 && mapping->nrexceptional == 0) + if (page_cache_empty(mapping)) return; invalidate_bh_lrus(); diff --git a/fs/dax.c b/fs/dax.c index 11b16729b86f..2f75ee2cd41f 100644 --- a/fs/dax.c +++ b/fs/dax.c @@ -949,7 +949,7 @@ int dax_writeback_mapping_range(struct address_space *mapping, if (WARN_ON_ONCE(inode->i_blkbits != PAGE_SHIFT)) return -EIO; - if (!mapping->nrexceptional || wbc->sync_mode != WB_SYNC_ALL) + if (page_cache_empty(mapping) || wbc->sync_mode != WB_SYNC_ALL) return 0; trace_dax_writeback_range(inode, xas.xa_index, end_index); diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h index 484a36185bb5..a474a92a2a72 100644 --- a/include/linux/pagemap.h +++ b/include/linux/pagemap.h @@ -18,6 +18,11 @@ struct pagevec; +static inline bool page_cache_empty(struct address_space *mapping) +{ + return xa_empty(&mapping->i_pages); +} + /* * Bits in mapping->flags. */ diff --git a/mm/truncate.c b/mm/truncate.c index dd9ebc1da356..7c4c8ac140be 100644 --- a/mm/truncate.c +++ b/mm/truncate.c @@ -300,7 +300,7 @@ void truncate_inode_pages_range(struct address_space *mapping, pgoff_t index; int i; - if (mapping->nrpages == 0 && mapping->nrexceptional == 0) + if (page_cache_empty(mapping)) goto out; /* Offsets within partial pages */ @@ -488,9 +488,6 @@ EXPORT_SYMBOL(truncate_inode_pages); */ void truncate_inode_pages_final(struct address_space *mapping) { - unsigned long nrexceptional; - unsigned long nrpages; - /* * Page reclaim can not participate in regular inode lifetime * management (can't call iput()) and thus can race with the @@ -500,16 +497,7 @@ void truncate_inode_pages_final(struct address_space *mapping) */ mapping_set_exiting(mapping); - /* - * When reclaim installs eviction entries, it increases - * nrexceptional first, then decreases nrpages. Make sure we see - * this in the right order or we might miss an entry. - */ - nrpages = mapping->nrpages; - smp_rmb(); - nrexceptional = mapping->nrexceptional; - - if (nrpages || nrexceptional) { + if (!page_cache_empty(mapping)) { /* * As truncation uses a lockless tree lookup, cycle * the tree lock to make sure any ongoing tree @@ -692,7 +680,7 @@ int invalidate_inode_pages2_range(struct address_space *mapping, int ret2 = 0; int did_range_unmap = 0; - if (mapping->nrpages == 0 && mapping->nrexceptional == 0) + if (page_cache_empty(mapping)) goto out; pagevec_init(&pvec);
Instead of checking the two counters (nrpages and nrexceptional), we can just check whether i_pages is empty. Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> --- fs/block_dev.c | 2 +- fs/dax.c | 2 +- include/linux/pagemap.h | 5 +++++ mm/truncate.c | 18 +++--------------- 4 files changed, 10 insertions(+), 17 deletions(-)