Whamcloud Community JIRA

Whamcloud Community JIRA https://jira.whamcloud.com This file is an XML representation of an issue en-us

9.4.14

940014

05-12-2023

[LU-12587] DIO fallback to Buffer IO unexpectedly https://jira.whamcloud.com/browse/LU-12587 Lustre <p>It is possible that DirectIO might fallback to Buffer IO in some cases, but following<br/> DIO fallback to Buffer IO very easily:</p> <p>dd if=/dev/zero of=data bs=1M count=1<br/> dd iff=/dev/zero of=data bs=1M count=1 oflag=direct conv=notrunc</p> LU-12587

DIO fallback to Buffer IO unexpectedly

Bug Minor Reopened Unresolved Patrick Farrell Wang Shilong always_except Thu, 25 Jul 2019 06:57:20 +0000 Fri, 21 Oct 2022 18:39:41 +0000 0 8 <p>Wang Shilong (wshilong@ddn.com) uploaded a new patch: <a href="https://review.whamcloud.com/35610" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/35610</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-12587" title="DIO fallback to Buffer IO unexpectedly" class="issue-link" data-issue-key="LU-12587">LU-12587</a> llite: don't check vmpage refcount in ll_releasepage()<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: 1<br/> Commit: eeb3f3884e0db092742daf88fc54422e207e5270</p> <p>Oleg Drokin (green@whamcloud.com) merged in patch <a href="https://review.whamcloud.com/35610/" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/35610/</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-12587" title="DIO fallback to Buffer IO unexpectedly" class="issue-link" data-issue-key="LU-12587">LU-12587</a> llite: don't check vmpage refcount in ll_releasepage()<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: <br/> Commit: e59f0c9a245f91ad929fd31dcbcbee88bf7a11df</p> <p>Landed for 2.13</p> <p>This change introduced <a href="https://jira.whamcloud.com/browse/LU-14541" title="Memory reclaim caused a stale data read" class="issue-link" data-issue-key="LU-14541"><del>LU-14541</del></a>. The original check here is a bit off, so fixing this is not just a revert, but mostly. See <a href="https://review.whamcloud.com/47262" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/47262</a></p> <p>We will have to see if there's another way to handle this issue. It might be possible to do the flushing here (see <a href="https://review.whamcloud.com/35610/" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/35610/</a> ) inside Lustre rather than relying on the kernel to do it.</p> <p>eg, if we flushed pages before generic_file_direct_write:</p> <div class='table-wrap'> <table class='confluenceTable'><tbody> <tr> <td class='confluenceTd'>->generic_file_direct_write()</td> </tr> <tr> <td class='confluenceTd'>->filemap_write_and_wait_range()</td> </tr> <tr> <td class='confluenceTd'>->invalidate_inode_pages2_range()</td> </tr> <tr> <td class='confluenceTd'>->invalidate_complete_page2() If a page can not be invalidated, return 0 to fall back to buffered write.</td> </tr> <tr> <td class='confluenceTd'>->try_to_release_page()</td> </tr> <tr> <td class='confluenceTd'>->ll_releasepage() return 0 becuase of vmpage count is 4 > 3</td> </tr> <tr> <td class='confluenceTd'>->generic_file_buffered_write</td> </tr> </tbody></table> </div> <p>It may just be that Lustre can't release pages in the desired manner... This will require some more investigation.</p> <p>sanity test_277 is being skipped on master because of <a href="https://jira.whamcloud.com/browse/LU-14541" title="Memory reclaim caused a stale data read" class="issue-link" data-issue-key="LU-14541"><del>LU-14541</del></a>, but it should really be referencing this ticket.</p> Related LU-14541 LU-15815 Development Rank 1|i00k5j: Rank (Obsolete) 9223372036854775807 Severity