Skip to contentSkip to conversation

nfs: wrong function to increment NFSIOS_READPAGES/WRITEPAGES stat counters?

PATCH New Nicolas Iooss 1 participant 2 messages Mar 24, 2015 – Apr 16, 2015

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hello,

Since commit 5a254d08b086 ("nfs: replace nfs_add_stats with nfs_inc_stats when add one") [1], nfs_readpage and nfs_do_writepage use nfs_inc_stats to increment NFSIOS_READPAGES and NFSIOS_WRITEPAGES instead of nfs_add_stats.

However nfs_inc_stats is not similar as nfs_add_stats with parameter 1 because these functions work on distinct stats:
nfs_inc_stats increments stats from "enum nfs_stat_eventcounters" (in server->io_stats->events) and nfs_add_stats those from "enum nfs_stat_bytecounters" (in server->io_stats->bytes), according to their implementations in fs/nfs/iostat.h [2].

If I understand the code correctly, "nfs_inc_stats(inode, NFSIOS_READPAGES);" is in fact executed as "nfs_inc_stats(inode, NFSIOS_VFSACCESS);" and "nfs_inc_stats(inode, NFSIOS_WRITEPAGES);" as "nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);".

As this looks like a bug to me, I am reporting it in hope it could be fixed, for example by reverting 5a254d08b086. Of course, I may be wrong, in which case an explanation about what I missed in my analysis would be appreciated.

By the way, I found this while trying to compile LLVMLinux on current master branch of Linux. clang reported the following warnings:

fs/nfs/read.c:287:23: warning: implicit conversion from enumeration type 'enum nfs_stat_bytecounters' to different enumeration type 'enum nfs_stat_eventcounters' [-Wenum-conversion]

            nfs_inc_stats(inode, NFSIOS_READPAGES);
            ~~~~~~~~~~~~~        ^~~~~~~~~~~~~~~~

fs/nfs/write.c:583:23: warning: implicit conversion from enumeration type 'enum nfs_stat_bytecounters' to different enumeration type 'enum nfs_stat_eventcounters' [-Wenum-conversion]

            nfs_inc_stats(inode, NFSIOS_WRITEPAGES);
            ~~~~~~~~~~~~~        ^~~~~~~~~~~~~~~~~

Thanks,

Nicolas

[1]
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=5a254d08b086d80cbead2ebcee6d2a4b3a15587a
[2]
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/fs/nfs/iostat.h?id=90a5a895cc8b284ac522757a01de15e36710c2b9

#1 lore raw

PATCH 1/1 Revert "nfs: replace nfs_add_stats with nfs_inc_stats when add one"

This reverts commit 5a254d08b086d80cbead2ebcee6d2a4b3a15587a.

Since commit 5a254d08b086 ("nfs: replace nfs_add_stats with nfs_inc_stats when add one"), nfs_readpage and nfs_do_writepage use nfs_inc_stats to increment NFSIOS_READPAGES and NFSIOS_WRITEPAGES instead of nfs_add_stats.

However nfs_inc_stats does not do the same thing as nfs_add_stats with value 1 because these functions work on distinct stats:
nfs_inc_stats increments stats from "enum nfs_stat_eventcounters" (in server->io_stats->events) and nfs_add_stats those from "enum nfs_stat_bytecounters" (in server->io_stats->bytes).

---

As I haven't got any reply from the message I sent a few weeks ago, I am now sending a patch.
More details about the reason of this revert can be found in my previous e-mail, https://lkml.org/lkml/2015/3/24/226.

 fs/nfs/read.c  | 2 +-
 fs/nfs/write.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
fs/nfs/read.c +1 -1
index 568ecf0a880f..848d8b1db4ce 100644
@@ -284,7 +284,7 @@ int nfs_readpage(struct file *file, struct page *page)
 	dprintk("NFS: nfs_readpage (%p %ld@%lu)\n",
 		page, PAGE_CACHE_SIZE, page_file_index(page));
 	nfs_inc_stats(inode, NFSIOS_VFSREADPAGE);
-	nfs_inc_stats(inode, NFSIOS_READPAGES);
+	nfs_add_stats(inode, NFSIOS_READPAGES, 1);
 
 	/*
 	 * Try to flush any pending writes to the file..
fs/nfs/write.c +1 -1
index 759931088094..fae7f97ad34d 100644
@@ -580,7 +580,7 @@ static int nfs_do_writepage(struct page *page, struct writeback_control *wbc, st
 	int ret;
 
 	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
-	nfs_inc_stats(inode, NFSIOS_WRITEPAGES);
+	nfs_add_stats(inode, NFSIOS_WRITEPAGES, 1);
 
 	nfs_pageio_cond_complete(pgio, page_file_index(page));
 	ret = nfs_page_async_flush(pgio, page, wbc->sync_mode == WB_SYNC_NONE);
Signature · 1 line
2.3.5