Linux NFS development
 help / color / mirror / Atom feed
From: Trond Myklebust <trondmy@hammerspace.com>
To: "bcodding@redhat.com" <bcodding@redhat.com>,
	"anna.schumaker@netapp.com" <anna.schumaker@netapp.com>
Cc: "linux-nfs@vger.kernel.org" <linux-nfs@vger.kernel.org>
Subject: Re: [PATCH] NFS: Always return the error that truncates a flushing page
Date: Mon, 28 Jan 2019 17:59:33 +0000	[thread overview]
Message-ID: <9fc3d4ee7d8eafdfdb7b7c1397ecb25dd22e12b7.camel@hammerspace.com> (raw)
In-Reply-To: <1cbf359e173a9f93e7c858faa43e623dda3858df.1548602308.git.bcodding@redhat.com>

On Sun, 2019-01-27 at 10:19 -0500, Benjamin Coddington wrote:
> We can't have nfs_wb_page() truncate the page from the mapping if
> there's
> an error on the context without returning that error, because we may
> be in
> nfs_updatepage() holding the page and trying to update the
> request.  Not
> having any error returned means we'll proceed to create a new request
> and
> dereference the truncated page->mapping.
> 
> If we're going to remove the page, always return the error that
> signaled us
> to do so in nfs_page_async_flush().
> 
> Fixes: c373fff7bd25 ("NFSv4: Don't special case "launder"")
> Cc: stable@vger.kernel.org # v4.11
> Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
> ---
>  fs/nfs/write.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/nfs/write.c b/fs/nfs/write.c
> index 5a0bbf917a32..c274339176cc 100644
> --- a/fs/nfs/write.c
> +++ b/fs/nfs/write.c
> @@ -622,9 +622,11 @@ static int nfs_page_async_flush(struct
> nfs_pageio_descriptor *pgio,
>  	WARN_ON_ONCE(test_bit(PG_CLEAN, &req->wb_flags));
>  
>  	ret = 0;
> -	/* If there is a fatal error that covers this write, just exit
> */
> -	if (nfs_error_is_fatal_on_server(req->wb_context->error))
> +	/* If there is a fatal on server error on this context, just
> exit */
> +	if (nfs_error_is_fatal_on_server(req->wb_context->error)) {
> +		ret = req->wb_context->error;
>  		goto out_launder;
> +	}
>  
>  	if (!nfs_pageio_add_request(pgio, req)) {
>  		ret = pgio->pg_error;

Hi Ben

We were apparently both looking at the same code last week ☺. I have a
similar patch, but that also fixes a similar clobbering issue with the
nfs_error_is_fatal() error just a few lines further down.

Without the extra hunk, we could end up converting an interrupted call
into a 'successful' write.

Cheers
  Trond

---
From 676d3340abc26ce9ff2b7609c293454dec6d4897 Mon Sep 17 00:00:00 2001
From: Trond Myklebust <trond.myklebust@hammerspace.com>
Date: Tue, 22 Jan 2019 07:34:45 -0500
Subject: [PATCH] NFS: Fix up return value on fatal errors in
 nfs_page_async_flush()

Ensure that we return the fatal error value that caused us to exit
nfs_page_async_flush().

Fixes: a6598813a4c5 ("NFS: Don't write back further requests...")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: stable@vger.kernel.org # v4.12+
---
 fs/nfs/write.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 5a0bbf917a32..f12cb31a41e5 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -621,11 +621,12 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
 	nfs_set_page_writeback(page);
 	WARN_ON_ONCE(test_bit(PG_CLEAN, &req->wb_flags));
 
-	ret = 0;
+	ret = req->wb_context->error;
 	/* If there is a fatal error that covers this write, just exit */
-	if (nfs_error_is_fatal_on_server(req->wb_context->error))
+	if (nfs_error_is_fatal_on_server(ret))
 		goto out_launder;
 
+	ret = 0;
 	if (!nfs_pageio_add_request(pgio, req)) {
 		ret = pgio->pg_error;
 		/*
@@ -635,9 +636,9 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
 			nfs_context_set_write_error(req->wb_context, ret);
 			if (nfs_error_is_fatal_on_server(ret))
 				goto out_launder;
-		}
+		} else
+			ret = -EAGAIN;
 		nfs_redirty_request(req);
-		ret = -EAGAIN;
 	} else
 		nfs_add_stats(page_file_mapping(page)->host,
 				NFSIOS_WRITEPAGES, 1);
-- 
2.20.1


-- 
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@hammerspace.com



  reply	other threads:[~2019-01-28 17:59 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-01-27 15:19 [PATCH] NFS: Always return the error that truncates a flushing page Benjamin Coddington
2019-01-28 17:59 ` Trond Myklebust [this message]
2019-01-28 18:55   ` Benjamin Coddington
2019-01-29 20:49     ` Trond Myklebust
2019-01-29 22:34       ` Benjamin Coddington
2019-01-30  9:03         ` Benjamin Coddington

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=9fc3d4ee7d8eafdfdb7b7c1397ecb25dd22e12b7.camel@hammerspace.com \
    --to=trondmy@hammerspace.com \
    --cc=anna.schumaker@netapp.com \
    --cc=bcodding@redhat.com \
    --cc=linux-nfs@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox