Skip to contentSkip to conversation

Basic reiserfs question

DISCUSSION Mack Stevenson 7 participants 12 messages Sep 6, 2001 – Sep 9, 2001

lore
Share & tools
lore
#0 lore raw
Quoted 18 lines
Quoted 2 lines · Mack Stevenson

I would like to know whether log replays take place whenever the system

is

Quoted 2 lines

booted or only after it has been incorrectly shutdown. I occasionaly see "Warning, log replay (...)" during the boot-up sequence although the

system

Quoted 1 line

had been correctly shut down, and I would like to know if I should

worry.

From: Hans Reiser <[email protected]>
To: Mack Stevenson <[email protected]>
CC: [email protected]
Subject: Re: Basic reiserfs question
Date: Thu, 06 Sep 2001 01:50:59 +0400

Quoted 2 lines

Are you using redhat? Do your shutdown scripts unmount the filesystem only if it is ext2?

I am using slackware (8.0), and the shutdown script on my system uses:

# Turn off swap, then unmount local file systems. echo "Turning off swap."
swapoff -a
echo "Unmounting local file systems."
# Don't remount UMSDOS root volumes:
if [ ! "mount | head -1 | cut -d ' ' -f 5" = "umsdos" ]; then
umount -a -r -t nonfs
echo "Remounting root filesystem read-only."
mount -n -o remount,ro /
else
umount -a -r -t nonfs,noumsdos,nosmbfs
fi

# This never hurts:
sync

---

I guess that's not the probem, then... Actually, whenever I shutdown I never see any messages about the unmounting (or the mounting in readonly mode) failing, which only adds to my confusion.

Quoted 8 lines

From: Samium Gromoff <[email protected]>
To: [email protected]
CC: [email protected]
Subject: Re: Basic reiserfs question
Date: Thu, 6 Sep 2001 03:35:49 +0000 (UTC)

Actually you see "Warning" only when it is replaying log on fs mounted readonly.

Yes, that's the warning I see.

Quoted 2 lines

I wander though why you see it only "occasionally".. Normally it replays log every time and you shouldnt bother.

Does it? I know next to nothing about these issues, but I thought that I had read somewhere that log replays only occurred after a system crashed and couldn't be shutdown properly.

Thank you both,

Mack S.

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp

#1 lore raw

It seems that we should put something in journal replay that says:

"Warning: replaying a non-empty journal, this means that either your system crashed, or its shutdown scripts need fixing (a common distro failing at the moment), or you pushed the power button. Don't use the hardware power button to turn your computer off before telling the operating system software to halt (there exists a 'halt' command you can use), the risk in doing so is that the files you or your software were writing to at the time you pushed the button can have garbage added to them."

Chris, do you agree? Edward, please make this change and create a patch.

Of course, now that the distros have turned off all those annoying messages from the kernel, most users will never see this, sigh. I think it is a real problem that the distros have turned off the messages at boot time....

Hans

Quoted 70 lines · Mack Stevenson
Quoted 18 lines
Quoted 2 lines · Mack Stevenson

I would like to know whether log replays take place whenever the system

is

Quoted 2 lines

booted or only after it has been incorrectly shutdown. I occasionaly see "Warning, log replay (...)" during the boot-up sequence although the

system

Quoted 1 line

had been correctly shut down, and I would like to know if I should

worry.

From: Hans Reiser <[email protected]>
To: Mack Stevenson <[email protected]>
CC: [email protected]
Subject: Re: Basic reiserfs question
Date: Thu, 06 Sep 2001 01:50:59 +0400

Quoted 3 lines

Are you using redhat? Do your shutdown scripts unmount the filesystem only if
it is ext2?

I am using slackware (8.0), and the shutdown script on my system uses:

# Turn off swap, then unmount local file systems. echo "Turning off swap."
swapoff -a
echo "Unmounting local file systems."
# Don't remount UMSDOS root volumes:
if [ ! "mount | head -1 | cut -d ' ' -f 5" = "umsdos" ]; then
umount -a -r -t nonfs
echo "Remounting root filesystem read-only."
mount -n -o remount,ro /
else
umount -a -r -t nonfs,noumsdos,nosmbfs
fi

# This never hurts:
sync

---

I guess that's not the probem, then... Actually, whenever I shutdown I never see any messages about the unmounting (or the mounting in readonly mode) failing, which only adds to my confusion.

Quoted 8 lines

From: Samium Gromoff <[email protected]>
To: [email protected]
CC: [email protected]
Subject: Re: Basic reiserfs question
Date: Thu, 6 Sep 2001 03:35:49 +0000 (UTC)

Actually you see "Warning" only when it is replaying log on fs mounted readonly.

Yes, that's the warning I see.

Quoted 2 lines

I wander though why you see it only "occasionally".. Normally it replays log every time and you shouldnt bother.

Does it? I know next to nothing about these issues, but I thought that I had read somewhere that log replays only occurred after a system crashed and couldn't be shutdown properly.

Thank you both,

Mack S.

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp

#2 lore raw
Quoted 11 lines · On Thursday, September 06, 2001 04:56:59 PM +0400 Hans Reiser…

It seems that we should put something in journal replay that says:

"Warning: replaying a non-empty journal, this means that either your system crashed, or its shutdown scripts need fixing (a common distro failing at the moment), or you pushed the power button. Don't use the hardware power button to turn your computer off before telling the operating system software to halt (there exists a 'halt' command you can use), the risk in doing so is that the files you or your software were writing to at the time you pushed the button can have garbage added to them."

Quoted 1 line

Chris, do you agree? Edward, please make this change and create a patch.

Sorry, I'm a bit slow this week, I caught a cold at linux world (apparently the users are contagious).

Anyway, the text above is great for a man page. kernel messages don't take more than one line ;-)

-chris

#3 lore raw
In reply to Mack Stevenson
Quoted 3 lines · Hans Reiser [[email protected]] wrote:

Of course, now that the distros have turned off all those annoying messages from the kernel, most users will never see this, sigh. I think it is a real problem that the distros have turned off the messages at boot time....

Heh, there are *so* many of them and they scroll by *so* fast. I'm not surprised they turned them off. They should just put a message at the very end "run the 'dmesg' command to read the preceeding boot messages.

Why is my cat pawing the desk next to my coffee like he's trying to cover my coffee with cat litter?!?!?! It's good coffee!!!

Cheers,
-- Bob

Bob McElrath ([email protected]) Univ. of Wisconsin at Madison, Department of Physics

#4 lore raw
In reply to Mack Stevenson

On Thu, 06 Sep 2001 16:56:59 +0400 Hans Reiser <[email protected]> wrote:

HR> It seems that we should put something in journal replay that says:
HR>
HR> "Warning: replaying a non-empty journal, this means that either your system HR> crashed, or its shutdown scripts need fixing (a common distro failing at the HR> moment)

If you think it's RedHat, you probably are wrong - I use RH with reiserfs a long time (more than a year - 6.2, now 7.1), and never got a message about replaying journal if system was shut down correctly.

Regards,
Nerijus

#5 lore raw

replaying reiserfs journal and bad blocks (was: Re[3]: Basic reiserfs question)

In reply to Mack Stevenson

NB> If you think it's RedHat, you probably are wrong - I use RH with reiserfs NB> a long time (more than a year - 6.2, now 7.1), and never got a message about NB> replaying journal if system was shut down correctly.

After just written that, I got a real problem. There was a power loss a few minutes ago, and I think when hdd was active. Trying to boot 2.4.7 kernel I got a message about replaying journal, then hdd error:

hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error }
hdd: dma_intr: error=0x40 { UncorrectableError }, LBAsect=84415, sector=36216

and kernel panic. Please note that I mount root read-write with this kernel.

Then I rebooted 2.2.19 kernel (which mounts root as read-only if that matters):

Checking ReiserFS transaction log (device 16:42) ...
Warning, log recovery starting on readonly filesystem
hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error }
hdd: dma_intr: error=0x40 { UncorrectableError }, LBAsect=84415, sector=36216
end_request: I/O error, dev 16:42 (hdd), sector 36216
Replayed 33 transactions in 6 seconds
Using r5 hash to sort names
...
ReiserFS version 3.5.32
VFS: Mounted root (reiserfs filesystem) readonly.

and system booted normally.
Now I rebooted back to 2.4.7 and everything seems to be OK. So what to do now? Run badblocks? Change disk? Any suggestions?

Regards,
Nerijus

#6 lore raw
In reply to Mack Stevenson
Quoted 13 lines · Nerijus Baliunas

On Thu, 06 Sep 2001 16:56:59 +0400 Hans Reiser <[email protected]> wrote:

HR> It seems that we should put something in journal replay that says:
HR>
HR> "Warning: replaying a non-empty journal, this means that either your system HR> crashed, or its shutdown scripts need fixing (a common distro failing at the HR> moment)

If you think it's RedHat, you probably are wrong - I use RH with reiserfs a long time (more than a year - 6.2, now 7.1), and never got a message about replaying journal if system was shut down correctly.

Quoted 8 lines

Regards,
Nerijus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

I only have secondhand reports from users who patch RedHat boot scripts as described at the end of www.namesys.com/faq.html, so your statement leaves me puzzled as to whether the secondhand reports were from persons who didn't understand the boot scripts. Comments are welcome.

Hans

#7 lore raw

Re: replaying reiserfs journal and bad blocks (was: Re[3]: Basic reiserfs question)

In reply to Mack Stevenson
Quoted 27 lines · Nerijus Baliunas

NB> If you think it's RedHat, you probably are wrong - I use RH with reiserfs NB> a long time (more than a year - 6.2, now 7.1), and never got a message about NB> replaying journal if system was shut down correctly.

After just written that, I got a real problem. There was a power loss a few minutes ago, and I think when hdd was active. Trying to boot 2.4.7 kernel I got a message about replaying journal, then hdd error:

hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error }
hdd: dma_intr: error=0x40 { UncorrectableError }, LBAsect=84415, sector=36216

and kernel panic. Please note that I mount root read-write with this kernel.

Then I rebooted 2.2.19 kernel (which mounts root as read-only if that matters):

Checking ReiserFS transaction log (device 16:42) ...
Warning, log recovery starting on readonly filesystem
hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error }
hdd: dma_intr: error=0x40 { UncorrectableError }, LBAsect=84415, sector=36216
end_request: I/O error, dev 16:42 (hdd), sector 36216
Replayed 33 transactions in 6 seconds
Using r5 hash to sort names
...
ReiserFS version 3.5.32
VFS: Mounted root (reiserfs filesystem) readonly.

Quoted 6 lines

and system booted normally.
Now I rebooted back to 2.4.7 and everything seems to be OK. So what to do now? Run badblocks? Change disk? Any suggestions?

Regards,
Nerijus

My guess is turn off UDMA, I think we have a www.namesys.com/faq.html entry on that which you can read and see if my memory of the typical symptoms of flaky udma are correct. Commenting on the cause of flaky udma I will leave to others familiar with your mob's interaction with Linux, except to say that one should always check cables at a time like this.

Hans

#8 lore raw

Re: replaying reiserfs journal and bad blocks (was: Re[3]: Basic

Quoted 2 lines
Quoted 2 lines

hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error }
hdd: dma_intr: error=0x40 { UncorrectableError }, LBAsect=84415, sector=36216

Thats the drive deciding it cant read the block.

Quoted 4 lines

My guess is turn off UDMA, I think we have a www.namesys.com/faq.html entry on
that which you can read and see if my memory of the typical symptoms of flaky
udma are correct. Commenting on the cause of flaky udma I will leave to others
familiar with your mob's interaction with Linux, except to say that one should> always check cables at a time like this.

You'd expect CRC errors then. It looks like the 2.2 log replay code worked but the 2.4 log replay code failed on the error, or the layers below it in the IDE got stuck.

#9 lore raw
In reply to Mack Stevenson

On Fri, 07 Sep 2001 11:50:06 +0400 Hans Reiser <[email protected]> wrote:

HR> Nerijus Baliunas wrote:
HR> >
HR> > On Thu, 06 Sep 2001 16:56:59 +0400 Hans Reiser <[email protected]> wrote:
HR> >
HR> > HR> It seems that we should put something in journal replay that says:
HR> > HR>
HR> > HR> "Warning: replaying a non-empty journal, this means that either your system HR> > HR> crashed, or its shutdown scripts need fixing (a common distro failing at the HR> > HR> moment)
HR> >
HR> > If you think it's RedHat, you probably are wrong - I use RH with reiserfs HR> > a long time (more than a year - 6.2, now 7.1), and never got a message about HR> > replaying journal if system was shut down correctly.
HR> >
HR> I only have secondhand reports from users who patch RedHat boot scripts as HR> described at the end of www.namesys.com/faq.html, so your statement leaves me HR> puzzled as to whether the secondhand reports were from persons who didn't HR> understand the boot scripts. Comments are welcome.

I didn't patch any boot scripts at all - they are pure RH ones.

Regards,
Nerijus

#10 lore raw
In reply to Mack Stevenson

On Fri, 07 Sep 2001 11:50:06 +0400 Hans Reiser <[email protected]> wrote:

HR> I only have secondhand reports from users who patch RedHat boot scripts as HR> described at the end of www.namesys.com/faq.html, so your statement leaves me HR> puzzled as to whether the secondhand reports were from persons who didn't HR> understand the boot scripts. Comments are welcome.

Hello,

I took time to read FAQ, and here are few comments.

1 - I understand you don't like RedHat, but instead of ranting, you could just say that gcc-2.96 from RH updates or RH 7.1 is OK (Alan Cox even recommends it for compiling recent 2.4 kernels).

16 - also "df -i" shows that there are no used inodes on reiserfs partitions.

36 - Can I use /sbin/badblocks -b <reiserfs-block-size> on existing reiserfs partition?

39 - a link to "I'm not using RedHat's special release of a buggy compiler"
(http://www.namesys.com/gcc-2.96) is not found.

40 - spelling error "ResierFS".

41 - there is nothing about how to solve it.

Regards,
Nerijus

#11 lore raw
In reply to Mack Stevenson
Quoted 6 lines · Nerijus Baliunas

HR> I only have secondhand reports from users who patch RedHat boot scripts as HR> described at the end of www.namesys.com/faq.html, so your statement leaves me HR> puzzled as to whether the secondhand reports were from persons who didn't HR> understand the boot scripts. Comments are welcome.

I didn't patch any boot scripts at all - they are pure RH ones.

It's true, that the RH 6.2 scripts only remount the root-FS read-only before rebooting if it is on ext2. This is only true for the root-FS (not any other FS), and I could check this only on RH 6.2 (Don't know about RH 7.x, I'd guess that they fixed it). If you don't have your root-FS on Reiser / XFS / JFS, you have no problem with RH 6.x

The FAQ entry from namesys is just misleading.

Regards
Henning

Signature · 5 lines
Dipl.-Inf. (Univ.) Henning P. Schmiedehausen       -- Geschaeftsfuehrer
INTERMETA - Gesellschaft fuer Mehrwertdienste mbH     [email protected]

Am Schwabachgrund 22  Fon.: 09131 / 50654-0   [email protected]
D-91054 Buckenhof     Fax.: 09131 / 50654-20