Skip to contentSkip to conversation

IDE-DMA woes

DISCUSSION Ishan Oshadi Jayawardena 7 participants 9 messages Dec 7, 2001 – Dec 10, 2001

lore
Share & tools
lore
#0 lore raw

Greetings.
I run Linux on an IBM PC300GL with Intel's
82371AB PIIX4 chipset. With DMA enabled (by doing a hdparm -d1 /dev/hda) on the hdd, I _sometimes_ get the following message from the kernel after resuming from APM standby mode:

ide_dmaproc: chipset supported ide_dma_timeout func only: 14
hda: status error: status=0x59 { DriveReady SeekComplete DataRequest Error }
hda: status error: error=0x84 { DriveStatusError BadCRC }
hda: drive not ready for command

then the drive stalls for a few seconds, and the driver disables DMA. This behaviour doesn't seem to depend on the kernel version (current: 2.4.14; error seen with 2.2 series also.) The weird thing is that this is not reliably reproducable; most of the time the system goes to apm standby (not suspend) and resumes fine.

Other (possibly relevent) boot-time messages (in order of appearance):

IBM machine detected. Enabling interrupts during APM calls.
kernel: PIIX4: IDE controller on PCI bus 00 dev 11
PIIX4: chipset revision 1
PIIX4: not 100%% native mode: will probe irqs later
ide0: BM-DMA at 0xfff0-0xfff7, BIOS settings: hda:DMA, hdb:pio
ide1: BM-DMA at 0xfff8-0xffff, BIOS settings: hdc:DMA, hdd:pio
hda: QUANTUM FIREBALL EX3.2A, ATA DISK drive
hdc: SONY CD-ROM CDU5221, ATAPI CDROM drive
ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
ide1 at 0x170-0x177,0x376 on irq 15

#1 lore raw
Quoted 6 lines · Ishan Oshadi Jayawardena

Greetings.
I run Linux on an IBM PC300GL with Intel's
82371AB PIIX4 chipset. With DMA enabled (by doing a hdparm -d1 /dev/hda) on the hdd, I _sometimes_ get the following message from the kernel after resuming from APM standby mode:

I have very similar behavior on an IBM Thinkpad T23. It's got this IDE controller:

00:1f.1 IDE interface: Intel Corporation: Unknown device 248a (rev 01)
Subsystem: IBM: Unknown device 0220

And, I also only sometimes get roughly:

ide_dmaproc: chipset supported ide_dma_lostirq func only: 13
hda: lost interrupt
ide_dmaproc: chipset supported ide_dma_timeout func only: 14

Quoted 11 lines

ide_dmaproc: chipset supported ide_dma_timeout func only: 14
hda: status error: status=0x59 { DriveReady SeekComplete DataRequest Error }
hda: status error: error=0x84 { DriveStatusError BadCRC }
hda: drive not ready for command

then the drive stalls for a few seconds, and the driver disables DMA. This behaviour doesn't seem to depend on the kernel version (current: 2.4.14; error seen with 2.2 series also.) The weird thing is that this is not reliably reproducable; most of the time the system goes to apm standby (not suspend) and resumes fine.

Unfortunately, when mine hits this condition, it seems to never recover from it. It also seems to only happen sometimes and I've been unable to reliably reproduce the problem. I told Andre about the problem and he suggested doing a "hdparm -d0 -X08 /dev/hda" prior to suspend and that seems to work around the problem. I "hdparm -d1 -X69 /dev/hda" on resume to get it back to speedy udma5 mode. I think the problem is the BIOS doing things to the IDE chipset during the suspend, and the driver not properly correcting the changes on resume.

-- Gerald

#2 lore raw

Gerald,

Like I told you in the other forum. I have the solutions just my work is being refused. Don't know if I have absolutely pissed off the king-penguin or what but there is no reason to submit when I know it is not going to be accepted -- regardless that it is in exact compliance to the ANSI/NCITS rules that I am now one of the co-authors.

Regards,

Andre Hedrick
CEO/President, LAD Storage Consulting Group
Linux ATA Development
Linux Disk Certification Project

Quoted 49 lines · Gerald Britton
Quoted 6 lines · Ishan Oshadi Jayawardena

Greetings.
I run Linux on an IBM PC300GL with Intel's
82371AB PIIX4 chipset. With DMA enabled (by doing a hdparm -d1 /dev/hda) on the hdd, I _sometimes_ get the following message from the kernel after resuming from APM standby mode:

I have very similar behavior on an IBM Thinkpad T23. It's got this IDE controller:

00:1f.1 IDE interface: Intel Corporation: Unknown device 248a (rev 01)
Subsystem: IBM: Unknown device 0220

And, I also only sometimes get roughly:

ide_dmaproc: chipset supported ide_dma_lostirq func only: 13
hda: lost interrupt
ide_dmaproc: chipset supported ide_dma_timeout func only: 14

Quoted 11 lines

ide_dmaproc: chipset supported ide_dma_timeout func only: 14
hda: status error: status=0x59 { DriveReady SeekComplete DataRequest Error }
hda: status error: error=0x84 { DriveStatusError BadCRC }
hda: drive not ready for command

then the drive stalls for a few seconds, and the driver disables DMA. This behaviour doesn't seem to depend on the kernel version (current: 2.4.14; error seen with 2.2 series also.) The weird thing is that this is not reliably reproducable; most of the time the system goes to apm standby (not suspend) and resumes fine.

Unfortunately, when mine hits this condition, it seems to never recover from it. It also seems to only happen sometimes and I've been unable to reliably reproduce the problem. I told Andre about the problem and he suggested doing a "hdparm -d0 -X08 /dev/hda" prior to suspend and that seems to work around the problem. I "hdparm -d1 -X69 /dev/hda" on resume to get it back to speedy udma5 mode. I think the problem is the BIOS doing things to the IDE chipset during the suspend, and the driver not properly correcting the changes on resume.

-- Gerald

-
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/

#3 lore raw

Andre -- Would you be willing to point us to where we can get it and play with it? I'm sure there are a lot of people here who would like to see your work. I, for one, am not sure where to get it though. :(

-Jacob

Quoted 14 lines · Andre Hedrick

Gerald,

Like I told you in the other forum. I have the solutions just my work is being refused. Don't know if I have absolutely pissed off the king-penguin or what but there is no reason to submit when I know it is not going to be accepted -- regardless that it is in exact compliance to the ANSI/NCITS rules that I am now one of the co-authors.

Regards,

Andre Hedrick
CEO/President, LAD Storage Consulting Group
Linux ATA Development
Linux Disk Certification Project

#4 lore raw
Quoted 5 lines · Jacob Luna Lundberg

Andre -- Would you be willing to point us to where we can get it and play with it? I'm sure there are a lot of people here who would like to see your work. I, for one, am not sure where to get it though. :(

-Jacob

I'd be a willing guinea pig too, if I knew where to look.

François

Quoted 26 lines · Andre Hedrick
Quoted 10 lines

Gerald,

Like I told you in the other forum. I have the solutions just my work is being refused. Don't know if I have absolutely pissed off the king-penguin or what but there is no reason to submit when I know it is not going to be accepted -- regardless that it is in exact compliance to the ANSI/NCITS rules that I am now one of the co-authors.

Quoted 7 lines

Regards,

Andre Hedrick
CEO/President, LAD Storage Consulting Group
Linux ATA Development
Linux Disk Certification Project

-
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/

#5 lore raw
In reply to Jacob Luna Lundberg

Me Too!

cu

jjs

Quoted 44 lines · François Cami
Quoted 5 lines · Jacob Luna Lundberg

Andre -- Would you be willing to point us to where we can get it and play with it? I'm sure there are a lot of people here who would like to see your work. I, for one, am not sure where to get it though. :(

-Jacob

I'd be a willing guinea pig too, if I knew where to look.

François

Quoted 19 lines · Andre Hedrick
Quoted 10 lines

Gerald,

Like I told you in the other forum. I have the solutions just my work is being refused. Don't know if I have absolutely pissed off the king-penguin or what but there is no reason to submit when I know it is not going to be accepted -- regardless that it is in exact compliance to the ANSI/NCITS rules that I am now one of the co-authors.

Quoted 7 lines

Regards,

Andre Hedrick
CEO/President, LAD Storage Consulting Group
Linux ATA Development
Linux Disk Certification Project

Quoted 7 lines

-
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/

-
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/

#6 lore raw
In reply to Jacob Luna Lundberg
Quoted 5 lines

Like I told you in the other forum. I have the solutions just my work is being refused. Don't know if I have absolutely pissed off the king-penguin or what but there is no reason to submit when I know it is not going to be accepted

Andre,
Any chance of some further details on this work. AFAIKT you've been the
foremost contributor of cutting-edge IDE updates for quite some time now. I find it astonishing that Linus would have inexplicably started refusing your code. At least point us in the right direction so we can try the new stuff out.

Simon

#7 lore raw
Quoted 6 lines · Simon Turvey
Quoted 5 lines

Like I told you in the other forum. I have the solutions just my work is being refused. Don't know if I have absolutely pissed off the king-penguin or what but there is no reason to submit when I know it is not going to be accepted

Quoted 8 lines

Andre,
Any chance of some further details on this work. AFAIKT you've been the
foremost contributor of cutting-edge IDE updates for quite some time now. I find it astonishing that Linus would have inexplicably started refusing your code. At least point us in the right direction so we can try the new stuff out.

Simon

I am working on a final update right now. This is after fixing the driver to perform correct data-transport layers. This is linux soft-raid 0 w/ four drives. The are 4-20GB drives partitioned to make two md devices to isolate peformance boundaries.

Writing superblocks and filesystem accounting information: done No size specified, using 510 MB
Size is MB, BlkSz is Bytes, Read, Write, and Seeks are MB/sec

         File   Block  Num  Seq Read    Rand Read   Seq Write  Rand Write
  Dir    Size   Size   Thr Rate (CPU%) Rate (CPU%) Rate (CPU%) Rate (CPU%)
------- ------ ------- --- ----------- ----------- ----------- -----------
   .     510    4096    1  84.75 57.1% 1.081 1.17% 79.80 43.9% 4.285 2.46%
   .     510    4096    2  78.88 52.8% 1.171 0.97% 73.64 48.8% 4.411 3.67%
   .     510    4096    4  74.04 51.3% 1.402 1.16% 73.77 51.9% 4.391 3.65%
   .     510    4096    8  67.93 47.9% 1.623 1.42% 71.43 52.1% 4.360 3.62%

File './Bonnie.943', size: 1073741824, volumes: 1

Writing with putc()...  done:   9709 kB/s  98.9 %CPU
Rewriting...            done:  49883 kB/s  41.4 %CPU
Writing intelligently...done:  99026 kB/s  50.0 %CPU
Reading with getc()...  done:   9663 kB/s  99.1 %CPU
Reading intelligently...done:  89632 kB/s  55.0 %CPU
Seeker 1...Seeker 2...Seeker 3...start 'em...done...done...done...
              ---Sequential Output (nosync)--- ---Sequential Input-- --Rnd Seek-
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --04k (03)-
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU   /sec %CPU
       1*1024  9709 98.9 99026 50.0 49883 41.4  9663 99.1 89632 55.0  344.2  4.0

Wrote 38184 Meg / 78201343 blocks 0 Meg / 0 blocks
Device length: 40038825984 Bytes / 38184 Meg / 37 Gig
Total Diameter Sequential Pattern Write Test = 96.10 MB/Sec (397.35 Seconds) Read 38183Meg (78198784 blocks)
Read failed on block number 78200832 at offset 40038825984 ( 0 / 512 )
Error: Input/output error
Could not find block: 78200833 for reading.
Error: Invalid argument
Read 38184Meg (78200833 blocks)
Device length: 40038825984 Bytes / 38184 Meg / 37 Gig
Total Diameter Sequential Pattern Read Test = 75.68 MB/Sec (504.53 Seconds) Device passed! Seek OverRun!

The test has an overrun on the last request under lseek, working on fixing.

The slower read rate is a direct result of using a pattern buffer check and comparison (write/read/verify/compare/contest-errors). Details later but the drive(s) are faster than reported. These rates are using EXT2 for the FS.

Regards,

Andre Hedrick
CEO/President, LAD Storage Consulting Group
Linux ATA Development
Linux Disk Certification Project

#8 lore raw
In reply to Ishan Oshadi Jayawardena
Quoted 9 lines · Gerald Britton

reliably reproduce the problem. I told Andre about the problem and he suggested doing a "hdparm -d0 -X08 /dev/hda" prior to suspend and that seems to work around the problem. I "hdparm -d1 -X69 /dev/hda" on resume to get it back to speedy udma5 mode. I think the problem is the BIOS doing things to the IDE chipset during the suspend, and the driver not properly correcting the changes on resume.

-- Gerald

Right. Thanks. I'll try it (hdparm -X66 actually, my drive only
supports udma2). (I know that it will work. I just want to know what's _exactly_ wrong with my setup ;)

-ioj
.