https://jira.whamcloud.com This file is an XML representation of an issue en-us 9.4.14 940014 05-12-2023 https://jira.whamcloud.com/browse/LU-2748 Lustre <div class="code panel" style="border-width: 1px;"><div class="codeContent panelContent"> <pre class="code-java">05:34:20:ll_ost_io00_029: page allocation failure. order:8, mode:0x50 05:34:20:Pid: 13919, comm: ll_ost_io00_029 Not tainted 2.6.32-279.19.1.el6_lustre.gc7121e9.x86_64 #1 05:34:20:Call Trace: 05:34:20: [&lt;ffffffff811231ff&gt;] ? __alloc_pages_nodemask+0x77f/0x940 05:34:20: [&lt;ffffffff8115d1a2&gt;] ? kmem_getpages+0x62/0x170 05:34:20: [&lt;ffffffff8115ddba&gt;] ? fallback_alloc+0x1ba/0x270 05:34:20: [&lt;ffffffff8115d80f&gt;] ? cache_grow+0x2cf/0x320 05:34:20: [&lt;ffffffff8115db39&gt;] ? ____cache_alloc_node+0x99/0x160 05:34:20: [&lt;ffffffffa04d8b60&gt;] ? cfs_alloc+0x30/0x60 [libcfs] 05:34:20: [&lt;ffffffff8115e909&gt;] ? __kmalloc+0x189/0x220 05:34:20: [&lt;ffffffffa04d8b60&gt;] ? cfs_alloc+0x30/0x60 [libcfs] 05:34:20: [&lt;ffffffffa0d1c8ae&gt;] ? osd_key_init+0x1e/0x670 [osd_ldiskfs] 05:34:20: [&lt;ffffffffa06643ef&gt;] ? keys_fill+0x6f/0x190 [obdclass] 05:34:20: [&lt;ffffffffa066843b&gt;] ? lu_context_init+0xab/0x260 [obdclass] 05:34:20: [&lt;ffffffffa07f78b3&gt;] ? ptlrpc_main+0x203/0x1870 [ptlrpc] 05:34:20: [&lt;ffffffff810097cc&gt;] ? __switch_to+0x1ac/0x320 05:34:20: [&lt;ffffffffa07f76b0&gt;] ? ptlrpc_main+0x0/0x1870 [ptlrpc] 05:34:20: [&lt;ffffffff8100c0ca&gt;] ? child_rip+0xa/0x20 05:34:20: [&lt;ffffffffa07f76b0&gt;] ? ptlrpc_main+0x0/0x1870 [ptlrpc] 05:34:20: [&lt;ffffffffa07f76b0&gt;] ? ptlrpc_main+0x0/0x1870 </pre> </div></div> <p>instead of kmalloc, kmem_cache_alloc should be used to allocate and effectively handle key's. </p> maloo LU-2748

OSD uses kmalloc with high order to allocate a keys

Technical task LU-1431 Minor Resolved Fixed Alex Zhuravlev Alexey Lyashkov Mon, 4 Feb 2013 13:53:59 +0000 Sat, 29 Jan 2022 08:34:07 +0000 Mon, 13 May 2013 21:50:02 +0000 Lustre 2.4.0 Lustre 2.4.0 0 10 <p>one more bug in that area.<br/> 4382 static void *osd_key_init(const struct lu_context *ctx,<br/> 4383 struct lu_context_key *key)<br/> ...<br/> 4391 OBD_ALLOC(info-&gt;oti_it_ea_buf, OSD_IT_EA_BUFSIZE);</p> <p>so 1Mb allocation via kmalloc, but for such buffer vmalloc prefered aka cfs_alloc_large.</p> <p>There are a number of problems found with PTLRPC_MAX_BRW_SIZE of 32MB that can be fixed relatively easily:</p> <ul class="alternate" type="square"> <li>osd_thread_info.osd_iobuf.dr_blocks[] is 512kB</li> <li>osd_thread_info.osd_iobuf.dr_pages[] is 64kB</li> <li>osd_thread_info.osd_iobuf is only needed for OST_IO_PORTAL and does not need to be allocated for other threads</li> <li>osd_thread_info.oti_created[] is 32kB and is <em>unused</em> and can be removed</li> <li>oti_thread_info uses OBD_ALLOC() instead of OBD_ALLOC_LARGE()</li> </ul> <p>There may be others, but these were obviously apparent at the first glance into the failed allocation paths.</p> <p>Probably dr_blocks[] can be reduced in size as well, since we will normally have 4kB blocksize instead of 1kB blocksize, so we don't need a buffer_head for every 1kB of space.</p> <p>Bandaid patch to just change the allocation to OBD_ALLOC_LARGE is in <a href="http://review.whamcloud.com/5323" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/5323</a>, this seems to greatly help in my testing.</p> <p><a href="http://review.whamcloud.com/#change,5444" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/#change,5444</a></p> <p>As per Oleg, dropping priority for remaining work</p> <p>I found a bug in <a href="http://review.whamcloud.com/#change,5444" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/#change,5444</a>. </p> <p>The rhel6.3 ldiskfs patch ext4-map_inode_page-2.6.18.patch defines ext4_map_inode_page function has having a integer array created need for it function but in fsfilt_ext3.c use call out to this function with</p> <p>extern int ext3_map_inode_page(struct inode *inode, struct page *page,<br/> unsigned long *blocks, int create);</p> <p>This will crash in a very painful way.</p> <p>Patch to fix this is at <a href="http://review.whamcloud.com/#change,5708" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/#change,5708</a></p> <p>Patch to fix this is at <a href="http://review.whamcloud.com/#change,5708" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/#change,5708</a></p> <p>should be resolved now?</p> <p>Oleg, can your <a href="http://review.whamcloud.com/#change,5323" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/#change,5323</a> "band aid" patch be abandoned?</p> <p>The <a href="http://review.whamcloud.com/5444" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/5444</a> patch was landed for 2.4.0, and <a href="http://review.whamcloud.com/5323" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/5323</a> is no longer needed.</p> Duplicate LU-2964 Related LU-2791 LU-2790 LU-1431 LU-2790 LU-2818 Development Rank 1|hzvijz: Rank (Obsolete) 6674