Bug 8816
| Summary: | ST340823A IDE disk problems with ide-disk.c, libata-core.c | ||
|---|---|---|---|
| Product: | IO/Storage | Reporter: | Mikko Rapeli (mikko.rapeli) |
| Component: | IDE | Assignee: | Bartlomiej Zolnierkiewicz (bzolnier) |
| Status: | RESOLVED PATCH_ALREADY_AVAILABLE | ||
| Severity: | normal | ||
| Priority: | P1 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Kernel Version: | 2.6.22.1 | Subsystem: | |
| Regression: | --- | Bisected commit-id: | |
| Attachments: |
kernel config for 2.6.22.1
dmesg with Debian kernel 2.6.8-4 dmesg with Debian kernel 2.6.10-1 dmesg with with 2.6.22.1 lshw output smartctl -a /dev/hdd, the drive seems ok dirty but working fix for 2.6.22.1 dmesg with 2.6.22.1 when drivers/ide/* compiled with -DDEBUG libata log with 2.6.22.1 libsata log with 2.6.22.1 and hpa disable by default (static int ata_ignore_hpa = 1;) blacklist hpa buggy devices in libata with 2.6.22.1 log with 2.6.22.1 plus hpa enabled by default and the previous blacklist patch Patch against 2.6.23-rc5 to add ST320413A to the black list in ide-disk.c |
||
Created attachment 12178 [details]
kernel config for 2.6.22.1
Created attachment 12179 [details]
dmesg with Debian kernel 2.6.8-4
Created attachment 12180 [details]
dmesg with Debian kernel 2.6.10-1
Created attachment 12181 [details]
dmesg with with 2.6.22.1
Created attachment 12182 [details]
lshw output
Created attachment 12213 [details]
smartctl -a /dev/hdd, the drive seems ok
Created attachment 12214 [details] dirty but working fix for 2.6.22.1 This disables the old stroke feature, what ever it did. Linus suggested this as a dirty fix for silent reboot hangs that were reported to exists between 2.6.10 and 2.6.14, http://marc.info/?l=linux-kernel&m=113060651817474&w=2 . The hang is very similar to what I described on the debian kernels between 2.6.12 and 2.6.16. Created attachment 12235 [details]
dmesg with 2.6.22.1 when drivers/ide/* compiled with -DDEBUG
Discussion about this bug: http://marc.info/?t=118596132600016&r=1&w=2 And the fix for 2.6.23-rc1: http://marc.info/?l=linux-ide&m=118609784802744&w=2 (Though 2.6.18 and 2.6.22.1 would propably like http://marc.info/?l=linux-ide&m=118609784802744&w=2 better.) Created attachment 12251 [details]
libata log with 2.6.22.1
Created attachment 12252 [details]
libsata log with 2.6.22.1 and hpa disable by default (static int ata_ignore_hpa = 1;)
Created attachment 12253 [details]
blacklist hpa buggy devices in libata with 2.6.22.1
Tested and gives a lot less warnings than normal hpa enabled kernel, though
the lvm partitions even without this patch as the previous log shows.
Created attachment 12254 [details]
log with 2.6.22.1 plus hpa enabled by default and the previous blacklist patch
Created attachment 12803 [details] Patch against 2.6.23-rc5 to add ST320413A to the black list in ide-disk.c ST320413A has the same problem as ST340823A. Please see https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.15/+bug/26119 Thanks. |
Most recent kernel where this bug did not occur: 2.6.8-4 from Debian sarge with smaller size, 2.6.12-1 from Debian with larger size, Debian kernels 2.6.13-1, 2.6.14-2 hang silently during boot at the point where HPA messages occur with older kernels, Debian kernels 2.6.16-2, 2.6.18-4 and 2.6.22.1 from kernel.org fail loudly Distribution: Debian etch Hardware Environment: *-ide:1 description: IDE Channel 1 physical id: 1 bus info: ide@1 logical name: ide1 clock: 33MHz *-disk description: ATA Disk product: ST340823A vendor: Seagate physical id: 1 bus info: ide@1.1 logical name: /dev/hdd version: 3.07 serial: 7EF1EYYG size: 37GB capacity: 37GB capabilities: ata dma lba iordy smart security pm partitioned partitioned:dos configuration: mode=udma2 smart=on *-volume:0 description: Linux LVM Physical Volume partition physical id: 1 bus info: ide@1.1,1 logical name: /dev/hdd1 serial: YwheMf-s0pv-97XS-sAOB-oGSk-WcHA-ZEo8Ag size: 9601MB capacity: 9601MB capabilities: primary multi lvm2 *-volume:1 description: Linux LVM Physical Volume partition physical id: 2 bus info: ide@1.1,2 logical name: /dev/hdd2 serial: TEN6QM-MVlA-LDlO-Moiu-O7mU-VTfZ-pinLz6 size: 15GB capacity: 15GB capabilities: primary multi lvm2 *-volume:2 description: Linux LVM Physical Volume partition physical id: 3 bus info: ide@1.1,3 logical name: /dev/hdd3 serial: b9Xviu-XSOP-9lz4-4gcY-tE9W-fqxA-k2amBh size: 12GB capacity: 12GB capabilities: primary multi lvm2 Software Environment: Debian etch with 2.6.22.1 kernel Problem Description: ST340823A IDE drive ( http://www.seagate.com/support/disc/ata/st340823a.html ) does not work with Debian kernels after 2.6.12-1 and with kernel.org 2.6.22.1: Probing IDE interface ide1... hdd: ST340823A, ATA DISK drive hdd: selected mode 0x42 ide1 at 0x170-0x177,0x376 on irq 15 ... hdd: max request size: 128KiB hdd: Host Protected Area detected. current capacity is 78165360 sectors (40020 MB) native capacity is 78165361 sectors (40020 MB) hdd: Host Protected Area disabled. hdd: 78165361 sectors (40020 MB) w/1024KiB Cache, CHS=65535/16/63, UDMA(33) hdd: cache flushes not supported hdd: hdd1 hdd2 hdd3 hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error } hdd: dma_intr: error=0x10 { SectorIdNotFound }, LBAsect=78165360, sector=78165360 ide: failed opcode was: unknown hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error } hdd: dma_intr: error=0x10 { SectorIdNotFound }, LBAsect=78165360, sector=78165360 ide: failed opcode was: unknown hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error } hdd: dma_intr: error=0x10 { SectorIdNotFound }, LBAsect=78165360, sector=78165360 ide: failed opcode was: unknown hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error } hdd: dma_intr: error=0x10 { SectorIdNotFound }, LBAsect=78165360, sector=78165360 ide: failed opcode was: unknown hdd: DMA disabled ide1: reset: master: error (0x00?) hdd: task_in_intr: status=0x59 { DriveReady SeekComplete DataRequest Error } hdd: task_in_intr: error=0x10 { SectorIdNotFound }, LBAsect=78230639, sector=78165360 ide: failed opcode was: unknown hdd: task_in_intr: status=0x59 { DriveReady SeekComplete DataRequest Error } hdd: task_in_intr: error=0x10 { SectorIdNotFound }, LBAsect=78230639, sector=78165360 ide: failed opcode was: unknown hdd: task_in_intr: status=0x59 { DriveReady SeekComplete DataRequest Error } hdd: task_in_intr: error=0x10 { SectorIdNotFound }, LBAsect=78230639, sector=78165360 ide: failed opcode was: unknown hdd: task_in_intr: status=0x59 { DriveReady SeekComplete DataRequest Error } hdd: task_in_intr: error=0x10 { SectorIdNotFound }, LBAsect=78230639, sector=78165360 ide: failed opcode was: unknown ide1: reset: master: error (0x00?) end_request: I/O error, dev hdd, sector 78165360 Buffer I/O error on device hdd, logical block 78165360 end_request: I/O error, dev hdd, sector 78165360 Buffer I/O error on device hdd, logical block 78165360 end_request: I/O error, dev hdd, sector 78165360 Buffer I/O error on device hdd, logical block 78165360 end_request: I/O error, dev hdd, sector 78165360 Buffer I/O error on device hdd, logical block 78165360 end_request: I/O error, dev hdd, sector 78165296 Buffer I/O error on device hdd, logical block 78165296 end_request: I/O error, dev hdd, sector 78165297 Buffer I/O error on device hdd, logical block 78165297 end_request: I/O error, dev hdd, sector 78165298 ... Disks filesystem is fsck clean under 2.6.8-4 and badblocks reports no errors with read and non-destructive write tests for the whole drive. Many users have been complaining about these drives and the problem seems to be the sector size, which Linux for some odd reason sees as being too big by one sector: https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.15/+bug/26119 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=401035 http://groups.google.fi/group/fa.linux.kernel/browse_thread/thread/68eaef2909603238/fd5574cfb1fb43a0?lnk=st&q=linux+ST340823A&rnum=13&hl=en&fwc=1 I think at least three factors contribute to this: 1) if kernel code is correct, disk reports two different sizes and larger one is incorrect, from dmesg on 2.6.8-4: current capacity is 78165360 sectors (40020 MB) native capacity is 78165361 sectors (40020 MB) Disk manual ( http://www.seagate.com/support/disc/manuals/ata/u5pmb01.pdf ) says that size n = 78165360 and that blocks 0 to n-1 are accessible with LBA mode, what ever it means. 2) kernels after 2.6.10 use the larger size as the disk size, diff between 2.6.8-4 and 2.6.10-1 Debian kernel dmesgs: +Probing IDE interface ide1... +ide1: Wait for ready failed before probe ! hdd: ST340823A, ATA DISK drive ide1 at 0x170-0x177,0x376 on irq 15 hdd: max request size: 128KiB hdd: Host Protected Area detected. current capacity is 78165360 sectors (40020 MB) native capacity is 78165361 sectors (40020 MB) -hdd: 78165360 sectors (40020 MB) w/1024KiB Cache, CHS=65535/16/63, UDMA(33) +hdd: Host Protected Area disabled. +hdd: 78165361 sectors (40020 MB) w/1024KiB Cache, CHS=65535/16/63, UDMA(33) +hdd: cache flushes not supported /dev/ide/host0/bus1/target1/lun0: p1 p2 p3 +Probing IDE interface ide2... +Probing IDE interface ide3... +Probing IDE interface ide4... +Probing IDE interface ide5... This alone does not seem to be a problem, although a file system accessing the last sector might wreck havoc, but at least I have not run into this. This size increase is due to "enable stroke by default" ( http://linux.bkbits.net:8080/linux-2.6.13-stable/?PAGE=cset&REV=1.1938.270.9 ) which has had some issues before ( http://marc.info/?l=linux-kernel&m=113060651817474&w=2 ). 3) Presumably, I'm speculating here, partition checking code starts reading the last sectors without error checking with 2.6.13 and with error checking after 2.6.16 which results in the disk failing loudly since the last sector does not exists while kernel thinks is should. 2.6.22.1 shouts: hdd: Host Protected Area detected. current capacity is 78165360 sectors (40020 MB) native capacity is 78165361 sectors (40020 MB) hdd: Host Protected Area disabled. hdd: 78165361 sectors (40020 MB) w/1024KiB Cache, CHS=65535/16/63, UDMA(33) hdd: cache flushes not supported hdd: hdd1 hdd2 hdd3 hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error } hdd: dma_intr: error=0x10 { SectorIdNotFound }, LBAsect=78165360, sector=78165360 ide: failed opcode was: unknown hdd: dma_intr: status=0x51 { DriveReady SeekComplete Error } hdd: dma_intr: error=0x10 { SectorIdNotFound }, LBAsect=78165360, sector=78165360 ide: failed opcode was: unknown ... Steps to reproduce: Try to use this drive with newer kernels than 2.6.12, presumably. Workaround(s): My first trial with 2.6.22.1 resulted in pata+libata being used and the drive worked fine, lvm partitions found and a clean filesystem too. After that 2.6.22.1 started to slow down to be unusable, but that may be a different issue. hdd=noprobe does not help and there seems to be no boot options with might help. Dumb questions: Why size increased by one sector between 2.6.8 and 2.6.10? stroke was enabled but why? http://linux.bkbits.net:8080/linux-2.6.13-stable/drivers/ide/ide-disk.c?PAGE=diffs&REV=1.108 http://linux.bkbits.net:8080/linux-2.6.13-stable/drivers/ide/ide-disk.c?PAGE=anno&REV=1.108 What's the difference between HPA current and native capacity? Does the drive report these current and native capacities correctly? How to force the smaller capacity if the drive reports it wrong? Why kernels after 2.6.12 access end of the disk? To support GUID/GDP/EFI partition type(s)? How to disable these scans? Could a filled up filesystem access the last sector and fail miserably even if partition scan didn't?