linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Andiry Xu <andiry@gmail.com>
To: Matthew Wilcox <willy@linux.intel.com>,
	Linux MM <linux-mm@kvack.org>,
	linux-kernel@vger.kernel.org
Cc: Andiry Xu <andiry.xu@gmail.com>
Subject: Filebench failure on ramdisk with Ext4-DAX
Date: Tue, 7 Jul 2015 12:19:48 -0700	[thread overview]
Message-ID: <CAOvWMLakCarg_4V9qPrG-TSUdqqBCBXMhJ3gHUXKNWf0Ym7YGQ@mail.gmail.com> (raw)

Hi,

I am running into failures when run filebench on ramdisk(/dev/ram0)
with Ext4-DAX.
The kernel version is 4.0, and I also verified it occurs on 4.2-rc1.

The issue reproduction steps:

// Set ramdisk size to 2GB
# mkfs.ext4 /dev/ram0
# mount -o dax /dev/ram0 /mnt/ramdisk
# filebench
filebench> load fileserver
filebench> set $dir=/mnt/ramdisk
filebench> run 30

And filebench fails in a few seconds like this:

8163: 22.992: Failed to pre-allocate file
/mnt/ramdisk/bigfileset/00000001/00000006/00000001/00000024/00000005/00000002/00000006:
No such file or directory on line 128
 8163: 22.992: Failed to create filesets on line 128

Or like this:

8141: 16.372: Failed to write 51967 bytes on fd 23: Success
 8151: 16.372: Failed to write 136735 bytes on fd 18: Success
 8148: 16.372: Failed to write 123317 bytes on fd 31: Success
 8141: 16.381: filereaderthread-36: flowop wrtfile1-1 failed
 8151: 16.381: filereaderthread-46: flowop wrtfile1-1 failed
 8148: 16.381: filereaderthread-43: flowop wrtfile1-1 failed
 8098: 16.521: Run took 1 seconds...
 8098: 16.521: NO VALID RESULTS! Filebench run terminated prematurely on line 65
 8098: 16.521: Shutting down processes

Sometimes it succeeds, but the chance is low. The failure rate is 80%+.

Note:
The issues does not occur with normal Ext4.
The issues does not occur with Ext4-DAX on pmem driver (from 01org/prd).

The only significant difference between brd.c and pmem.c is that brd.c
uses alloc_page() and pmem.c reserved memory range and uses ioremap()
to get virtual address. I assume that the memcpy
operation(copy_from/to_user) directly between user buffer and page by
alloc_page() does not work correctly somehow. I wonder if this is a
bug? If it is, how to fix it? Thanks.

Thanks,
Andiry

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

                 reply	other threads:[~2015-07-07 19:19 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

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=CAOvWMLakCarg_4V9qPrG-TSUdqqBCBXMhJ3gHUXKNWf0Ym7YGQ@mail.gmail.com \
    --to=andiry@gmail.com \
    --cc=andiry.xu@gmail.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=willy@linux.intel.com \
    /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