rpciod process is blocked in nfs_release_page waiting for nfs_commit_inode to complete

[Harshula <harshula@redhat.com>]

Hi All,

Can I please get your feedback on the following?

rpciod is blocked:
------------------------------------------------------------
crash> bt  2507 
PID: 2507   TASK: ffff88103691ab40  CPU: 14  COMMAND: "rpciod/14"
 #0 [ffff8810343bf2f0] schedule at ffffffff814dabd9
 #1 [ffff8810343bf3b8] nfs_wait_bit_killable at ffffffffa038fc04 [nfs]
 #2 [ffff8810343bf3c8] __wait_on_bit at ffffffff814dbc2f
 #3 [ffff8810343bf418] out_of_line_wait_on_bit at ffffffff814dbcd8
 #4 [ffff8810343bf488] nfs_commit_inode at ffffffffa039e0c1 [nfs]
 #5 [ffff8810343bf4f8] nfs_release_page at ffffffffa038bef6 [nfs]
 #6 [ffff8810343bf528] try_to_release_page at ffffffff8110c670
 #7 [ffff8810343bf538] shrink_page_list.clone.0 at ffffffff81126271
 #8 [ffff8810343bf668] shrink_inactive_list at ffffffff81126638
 #9 [ffff8810343bf818] shrink_zone at ffffffff8112788f
#10 [ffff8810343bf8c8] do_try_to_free_pages at ffffffff81127b1e
#11 [ffff8810343bf958] try_to_free_pages at ffffffff8112812f
#12 [ffff8810343bfa08] __alloc_pages_nodemask at ffffffff8111fdad
#13 [ffff8810343bfb28] kmem_getpages at ffffffff81159942
#14 [ffff8810343bfb58] fallback_alloc at ffffffff8115a55a
#15 [ffff8810343bfbd8] ____cache_alloc_node at ffffffff8115a2d9
#16 [ffff8810343bfc38] kmem_cache_alloc at ffffffff8115b09b
#17 [ffff8810343bfc78] sk_prot_alloc at ffffffff81411808
#18 [ffff8810343bfcb8] sk_alloc at ffffffff8141197c
#19 [ffff8810343bfce8] inet_create at ffffffff81483ba6
#20 [ffff8810343bfd38] __sock_create at ffffffff8140b4a7
#21 [ffff8810343bfd98] xs_create_sock at ffffffffa01f649b [sunrpc]
#22 [ffff8810343bfdd8] xs_tcp_setup_socket at ffffffffa01f6965 [sunrpc]
#23 [ffff8810343bfe38] worker_thread at ffffffff810887d0
#24 [ffff8810343bfee8] kthread at ffffffff8108dd96
#25 [ffff8810343bff48] kernel_thread at ffffffff8100c1ca
------------------------------------------------------------

nfs_release_page() gives kswapd process an exemption from being blocked.
Should we do the same for rpciod. Otherwise rpciod could end up in a
deadlock where it can not continue without freeing memory that will only
become available when rpciod does its work:
------------------------------------------------------------
479 /*
480  * Attempt to release the private state associated with a page
481  * - Called if either PG_private or PG_fscache is set on the page
482  * - Caller holds page lock
483  * - Return true (may release page) or false (may not)
484  */
485 static int nfs_release_page(struct page *page, gfp_t gfp)
486 {   
487         struct address_space *mapping = page->mapping;
488     
489         dfprintk(PAGECACHE, "NFS: release_page(%p)\n", page);
490     
491         /* Only do I/O if gfp is a superset of GFP_KERNEL */
492         if (mapping && (gfp & GFP_KERNEL) == GFP_KERNEL) {
493                 int how = FLUSH_SYNC;
494             
495                 /* Don't let kswapd deadlock waiting for OOM RPC calls */
496                 if (current_is_kswapd())
497                         how = 0;
498                 nfs_commit_inode(mapping->host, how);
499         }
500         /* If PagePrivate() is set, then the page is not freeable */
501         if (PagePrivate(page))
502                 return 0;
503         return nfs_fscache_release_page(page, gfp);
504 }
------------------------------------------------------------

Another option is to change the priority of the memory allocation:
net/ipv4/af_inet.c
------------------------------------------------------------
 265 static int inet_create(struct net *net, struct socket *sock, int
protocol,
 266                        int kern)
 267 {
...
 345         sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
------------------------------------------------------------
Considering this is generic net code, I assume the GFP_KERNEL will not
be replaced with GFP_ATOMIC.

NOTE, this is on RHEL 6.1 kernel 2.6.32-131.6.1 and apparently uses
'legacy' workqueue code.

cya,
#