History

mm/khugepaged: fix crashes due to misaccounted holes

Huge tmpfs testing on a shortish file mapped into a pmd-rounded extent hit shmem_evict_inode()'s WARN_ON(inode->i_blocks) followed by clear_inode()'s BUG_ON(inode->i_data.nrpages) when the file was later closed and unlinked. khugepaged's collapse_shmem() was forgetting to update mapping->nrpages on the rollback path, after it had added but then needs to undo some holes. There is indeed an irritating asymmetry between shmem_charge(), whose callers want it to increment nrpages after successfully accounting blocks, and shmem_uncharge(), when __delete_from_page_cache() already decremented nrpages itself: oh well, just add a comment on that to them both. And shmem_recalc_inode() is supposed to be called when the accounting is expected to be in balance (so it can deduce from imbalance that reclaim discarded some pages): so change shmem_charge() to update nrpages earlier (though it's rare for the difference to matter at all).

Hugh Dickins committed 7 years ago

Commit aaa52e31 parent 701270f2 files · +9 −2Browse files

Unified · Split

2 files
  1. mm/khugepaged.c +4 −1
  2. mm/shmem.c +5 −1
mm/khugepaged.c+4−1
@@ -1506,9 +1506,12 @@ xa_unlocked: · 1505 unmodified lines
1506 khugepaged_pages_collapsed++;1506 khugepaged_pages_collapsed++;
1507 } else {1507 } else {
1508 struct page *page;1508 struct page *page;
1509
1509 /* Something went wrong: roll back page cache changes */1510 /* Something went wrong: roll back page cache changes */
1510 shmem_uncharge(mapping->host, nr_none);
1511 xas_lock_irq(&xas);1511 xas_lock_irq(&xas);
1512 mapping->nrpages -= nr_none;
1513 shmem_uncharge(mapping->host, nr_none);
1514
1512 xas_set(&xas, start);1515 xas_set(&xas, start);
1513 xas_for_each(&xas, page, end - 1) {1516 xas_for_each(&xas, page, end - 1) {
1514 page = list_first_entry_or_null(&pagelist,1517 page = list_first_entry_or_null(&pagelist,
mm/shmem.c+5−1
@@ -297,12 +297,14 @@ bool shmem_charge(struct inode *inode, long pages) · 296 unmodified lines
297 if (!shmem_inode_acct_block(inode, pages))297 if (!shmem_inode_acct_block(inode, pages))
298 return false;298 return false;
299 299
300 /* nrpages adjustment first, then shmem_recalc_inode() when balanced */
301 inode->i_mapping->nrpages += pages;
302
300 spin_lock_irqsave(&info->lock, flags);303 spin_lock_irqsave(&info->lock, flags);
301 info->alloced += pages;304 info->alloced += pages;
302 inode->i_blocks += pages * BLOCKS_PER_PAGE;305 inode->i_blocks += pages * BLOCKS_PER_PAGE;
303 shmem_recalc_inode(inode);306 shmem_recalc_inode(inode);
304 spin_unlock_irqrestore(&info->lock, flags);307 spin_unlock_irqrestore(&info->lock, flags);
305 inode->i_mapping->nrpages += pages;
306 308
307 return true;309 return true;
308}310}
@@ -312,6 +314,8 @@ void shmem_uncharge(struct inode *inode, long pages) · 3 unmodified lines
312 struct shmem_inode_info *info = SHMEM_I(inode);314 struct shmem_inode_info *info = SHMEM_I(inode);
313 unsigned long flags;315 unsigned long flags;
314 316
317 /* nrpages adjustment done by __delete_from_page_cache() or caller */
318
315 spin_lock_irqsave(&info->lock, flags);319 spin_lock_irqsave(&info->lock, flags);
316 info->alloced -= pages;320 info->alloced -= pages;
317 inode->i_blocks -= pages * BLOCKS_PER_PAGE;321 inode->i_blocks -= pages * BLOCKS_PER_PAGE;

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t