Skip to contentSkip to conversation

Linux Kernel 2.6.6 IDE shutdown problems.

DISCUSSION Justin Piszcz 8 participants 10 messages May 15, 2004 – May 26, 2004

lore
Share & tools
lore
#0 lore raw

The problem is the 2.6.6 kernel muxed my drive and when it fscked upon reboot it deleted /etc/mtab and lilo.conf!

Luckily I restored them from a backup and now run 2.6.5 and it is working fine.

Linux 2.6.6 is a nightmare.

I am looking into the benchmark problem with 2.6.6 now.

Quoted 10 lines · Justin Piszcz

Now whenever I reboot it says input/output errors when it tries to mount the drive? I will look into this further.

_________________________________________________________________
Stop worrying about overloading your inbox - get MSN Hotmail Extra Storage! http://join.msn.click-url.com/go/onm00200362ave/direct/01/

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

_________________________________________________________________
Check out the coupons and bargains on MSN Offers! http://youroffers.msn.com

#1 lore raw
Quoted 2 lines · Justin Piszcz

The problem is the 2.6.6 kernel muxed my drive and when it fscked upon reboot it deleted /etc/mtab and lilo.conf!

What fs are you using?

Quoted 10 lines

Luckily I restored them from a backup and now run 2.6.5 and it is working fine.

Linux 2.6.6 is a nightmare.

I am looking into the benchmark problem with 2.6.6 now.

Quoted 2 lines · Justin Piszcz

Now whenever I reboot it says input/output errors when it tries to mount the drive? I will look into this further.

This errors are HARMLESS and CAN'T corrupt your data.
Please see http://bugme.osdl.org/show_bug.cgi?id=2672 for description+fix.

Cheers,
Bartlomiej

#2 lore raw
Quoted 16 lines · Bartlomiej Zolnierkiewicz
Quoted 2 lines · Justin Piszcz

The problem is the 2.6.6 kernel muxed my drive and when it fscked upon reboot it deleted /etc/mtab and lilo.conf!

What fs are you using?

Quoted 7 lines

Luckily I restored them from a backup and now run 2.6.5 and it is working fine.

Linux 2.6.6 is a nightmare.

I am looking into the benchmark problem with 2.6.6 now.

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines · Justin Piszcz

Now whenever I reboot it says input/output errors when it tries to mount the drive? I will look into this further.

This errors are HARMLESS and CAN'T corrupt your data.
Please see http://bugme.osdl.org/show_bug.cgi?id=2672 for description+fix.

I would think that if the drive didn't properly flush cache on shutdown that it might cause corruption. Feel free to tell me no drive would bahave like that ;-)

Signature · 3 lines
   -bill davidsen ([email protected])
"The secret to procrastination is to put things off until the
 last possible moment - but no longer"  -me
#3 lore raw
Quoted 3 lines · Bill Davidsen

I would think that if the drive didn't properly flush cache on shutdown that it might cause corruption. Feel free to tell me no drive would bahave like that ;-)

why not add a one or two second delay before? i doubt any drive holds its writeback that long.

Signature · 2 lines
Tom Vier <[email protected]>
DSA Key ID 0x15741ECE
#4 lore raw
Quoted 7 lines · Tom Vier
Quoted 3 lines · Bill Davidsen

I would think that if the drive didn't properly flush cache on shutdown that it might cause corruption. Feel free to tell me no drive would bahave like that ;-)

why not add a one or two second delay before? i doubt any drive holds its writeback that long.

This was the solution adopted by IBM, for their Thinkpad laptop line, in a BIOS update last year. My system now delays two seconds before turning off, with the stated purpose "avoiding disk corruption"...

Daniel

Signature · 2 lines
The English have the most rigid code of immorality in the world.
        -- Malcom Bradbury, _Eating People is Wrong_
#5 lore raw
In reply to Tom Vier
Quoted 5 lines · Tom Vier
Quoted 1 line · Bill Davidsen

I would think that if the drive didn't properly flush cache on shutdown that it might cause corruption. Feel free to tell me no drive would bahave like that ;-)

why not add a one or two second delay before? i doubt any drive holds its writeback that long.

That is an arbitrary delay that isn't guaranteed to work in all workloads.

FLUSH CACHE is the way to do this, complain to vendors that don't support flush cache and get them to fix their drives. (Like how Bart spoke to me and we both 1) worked out a solution for current drives and 2) fixed the root case in all future firmwares)

Picture a nice fast drive doing 100 writes/second to the media... if you give it over 200 writes at a time, it'll occupy your 2 seconds. Newer drives with 8MB or larger buffers are certainly capable of caching a lot more than 200 writes...

Drives are getting smarter so they can better work around these sorts of cases, but FLUSH CACHE (EXT) is the only way that should be guaranteed to work.

--eric

Signature · 2 lines
Eric D. Mudama
[email protected]
#6 lore raw
Quoted 4 lines · Eric D. Mudama

Picture a nice fast drive doing 100 writes/second to the media... if you give it over 200 writes at a time, it'll occupy your 2 seconds. Newer drives with 8MB or larger buffers are certainly capable of caching a lot more than 200 writes...

Quite unlikely. Usually disks have a big cache but it can hold a very limited number of blocks. 8MB of cache is probably divided in 8 blocks of 1MB each.

--
Giuliano.

#7 lore raw
Quoted 4 lines · Giuliano Pochini

On Mon, 24 May 2004, Eric D. Mudama wrote:

Quoted 8 lines
Quoted 4 lines

Picture a nice fast drive doing 100 writes/second to the media... if you give it over 200 writes at a time, it'll occupy your 2 seconds. Newer drives with 8MB or larger buffers are certainly capable of caching a lot more than 200 writes...

Quite unlikely. Usually disks have a big cache but it can hold a very limited number of blocks. 8MB of cache is probably divided in 8 blocks of 1MB each.

Sorry, but that isn't true, unless some company is just plain stupid.

Everyone has different metrics for cache granularity based on their cache architecture, but I can assure you that 8x 1MB segments is off by 1-2 orders of magnitude, and has been for years.

In practice, depending on the workload, there may appear to only be 8 active segments as drives today can merge cache segments or other similar things (architectually dependant), but the worst-case (best case?) is significantly more.

--eric

Signature · 2 lines
Eric D. Mudama
[email protected]
#8 lore raw
In reply to Giuliano Pochini
Quoted 9 lines · Giuliano Pochini
Quoted 4 lines · Eric D. Mudama

Picture a nice fast drive doing 100 writes/second to the media... if you give it over 200 writes at a time, it'll occupy your 2 seconds. Newer drives with 8MB or larger buffers are certainly capable of caching a lot more than 200 writes...

Quite unlikely. Usually disks have a big cache but it can hold a very limited number of blocks. 8MB of cache is probably divided in 8 blocks of 1MB each.

No.

#9 lore raw
Quoted 9 lines · Bartlomiej Zolnierkiewicz
Quoted 7 lines · Giuliano Pochini
Quoted 4 lines · Eric D. Mudama

Picture a nice fast drive doing 100 writes/second to the media... if you give it over 200 writes at a time, it'll occupy your 2 seconds. Newer drives with 8MB or larger buffers are certainly capable of caching a lot more than 200 writes...

Quoted 6 lines
Quoted 3 lines

Quite unlikely. Usually disks have a big cache but it can hold a very limited number of blocks. 8MB of cache is probably divided in 8 blocks of 1MB each.

No.

It is indeed likely that the worst case is around 16000 writes, which if on seperate tracks may take over 30 seconds to complete if done individually.

However, it's likely that any drive designer with a clue would allocate a couple of journal tracks, so that the write cache and two backup copies can be stored for replay when the drive is powered on again.

Why do I suspect that some designers don't have clue, or haven't really thought enough about this case.