HPT370 (KT7A-RAID) *corrupts* data - SAMSUNG SV8004H does it as well
This is an addition to my reports about HPT370 corrupting data with a pair of IBM-DPTA-373420's on linux. As a summary, my testing showed that hpt370 + IBM-DPTA-373420 corrupt data on
- 2.2.18pre19 + ide patch
- 2.2.20 + ide patch
- 2.2.20 + ide patch + Tim Hockin's hpt366.c patch
- 2.2.20 + ide patch + Tim Hockin's hpt366.c patch, in UDMA33 mode (rather than UDMA66)
- 2.4.15 + Tim Hockin's hpt366.c patch
The test involved reading /dev/md0 (that consists of /dev/hde and /dev/hdg) several times and comparing the md5sums. I also tried reading /dev/hde and /dev/hdg in parallel, and it did show errors. The problem disappeared when I moved the drives over to Via 868B interface.
I reported the problem to Highpoint Tech Inc as well (they do explicitly list IBM-DPTA-373420 as tested and compatible with HPT370), but as anticipated, they didn't answer.
Now I bought a pair of SAMSUNG SV8004H's. Since the drives were blank, I was able to do a write test. The test (see http://v.iki.fi/~vherva/tmp/wrchk.c
for the quick'n'dirty proggie) writes the /dev/md1 (which again consists of the two SAMSUNG SV8004H's) full of a certain randomish 64MB block and then tries to read it back. The write and read cycles are done over and over again.
This is with 2.2.20 + ide + Hockin's patch. Drives are in UDMA100 mode (the default) and no hdparm adjustions have been made.
The first write-read cycle went well, but one block mismatched already on the second run. I've only run the test over night, but there are already several mimatches (see http://v.iki.fi/~vherva/tmp/samsung-log for complete log.)
Even during the succesfull first run, the IDE system gave these warningins:
hdg: status error: status=0x58 { DriveReady SeekComplete DataRequest }
hdg: drive not ready for command
Later during the night, I got 27 of those errors. I also got 10 of these:
probable hardware bug: clock timer configuration lost - probably a VIA686a.
I didn't get either of these with the IBM disks.
Smartctl shows 'No Errors Logged' for both drives. Also, it reports the temperature of the drives being always under 35 degrees Celsius, at times even under 30. I reckon temperatur is not a problem.
Right now I'm wondering two things:
- how come anyone else is not seeing this corruption (Abit KT7A, nevermind HPT370 is fairly popular)?
- is it safe to solder the bugger off the motherboard so I can introduce it to my shotgun?
regards,
Signature · 3 lines
Ville Herva [email protected] +358-40-5756996 Viasys Oy Hannuntie 6 FIN-02360 Espoo +358-9-2313-2160 PGP key available: http://www.iki.fi/v/pgp.html fax +358-9-2313-2250
Quoted 2 lines · Ville Herva
- how come anyone else is not seeing this corruption (Abit KT7A, nevermind HPT370 is fairly popular)?
A friend of mine had an IBM DLTA drive attached to his HPT370 controller, and this combination proved to produce a whole lot of drive errors (I can confirm this first hand), which went away after attaching the drive to the main motherboard controller.
I can't say anything about data corruption though - I just asked him and he said he didn't know of any, but that doesn't mean it didn't happen.
Regs,
Sven
Signature · 4 lines
Sven Riedel [email protected] Osteroeder Str. 6 / App. 13 [email protected] 38678 Clausthal "Call me bored, but don't call me boring." - Larry Wall
On Sat, Dec 01, 2001 at 11:34:00AM +0100, you [[email protected]] claimed:
Quoted 10 lines · Ville Herva
Quoted 2 lines
- how come anyone else is not seeing this corruption (Abit KT7A, nevermind HPT370 is fairly popular)?
A friend of mine had an IBM DLTA drive attached to his HPT370 controller, and this combination proved to produce a whole lot of drive errors (I can confirm this first hand), which went away after attaching the drive to the main motherboard controller.
I can't say anything about data corruption though - I just asked him and he said he didn't know of any, but that doesn't mean it didn't happen.
Of course the drive is longer attached to HPT370 and your friend is propably reluctant to reattach it, but it would still be nice to know if he gets consistent results which for example this simple test:
cat /dev/hde | mdsum
run for several (5-10, perhaps) times.
OTOH, I haven't had corruption with reading only
one disk at a time, but then again I haven't tried too hard as they should really work in parallel.
-- v --
Quoted 3 lines · Ville Herva
Of course the drive is longer attached to HPT370 and your friend is propably reluctant to reattach it, but it would still be nice to know if he gets consistent results which for example this simple test:
Well, I asked him, but he doesn't want to rip his fileserver apart again. Sorry.
Regs,
Sven
Signature · 4 lines
Sven Riedel [email protected] Osteroeder Str. 6 / App. 13 [email protected] 38678 Clausthal "Call me bored, but don't call me boring." - Larry Wall
In article <[email protected]> you write:
Quoted 2 lines
- how come anyone else is not seeing this corruption (Abit KT7A, nevermind HPT370 is fairly popular)?
I've got the KT7-RAID (note, no A) but with only one drive attached to the
HPT370. I have seen no problems. I will try your recommended md5sum test next time I'm at the console to be root...
Jonathan.
On Tue, Dec 04, 2001 at 06:15:51PM +0000, you [Jonathan Amery] claimed:
Quoted 7 lines
In article <[email protected]> you write:
Quoted 2 lines
- how come anyone else is not seeing this corruption (Abit KT7A, nevermind HPT370 is fairly popular)?
I've got the KT7-RAID (note, no A) but with only one drive attached to the
HPT370. I have seen no problems. I will try your recommended md5sum test next time I'm at the console to be root...
I've only ever seen the corruption when reading two disks in parallel (one at each hpt ide channel).
-- v --
Setup here is:
hde: WDC WD200EB-00BHF0, ATA DISK drive
hdg: WDC WD200EB-00BHF0, ATA DISK drive
hde: 39102336 sectors (20020 MB) w/2048KiB Cache, CHS=38792/16/63, UDMA(100)
hdg: 39102336 sectors (20020 MB) w/2048KiB Cache, CHS=38792/16/63, UDMA(100)
5GB of each in RAID0 on /dev/md/2
cat /dev/md/2 | md5sum now done its fourth run; all OK. 920b175a519b578dcd7862b720eb9efb, if you care ;-)
This is 2.4.6, on a KT7-RAID board.
On Tue, Dec 04, 2001 at 05:23:19PM -0500, you [Anthony DeRobertis] claimed:
Quoted 10 lines
Setup here is:
hde: WDC WD200EB-00BHF0, ATA DISK drive
hdg: WDC WD200EB-00BHF0, ATA DISK drivehde: 39102336 sectors (20020 MB) w/2048KiB Cache,
CHS=38792/16/63, UDMA(100)
hdg: 39102336 sectors (20020 MB) w/2048KiB Cache,
CHS=38792/16/63, UDMA(100)
Quoted 4 lines
5GB of each in RAID0 on
/dev/md/2cat
/dev/md/2| md5sum now done its fourth run; all OK.920b175a519b578dcd7862b720eb9efb, if you care ;-)
I do care (not of the sum, but of the fact your sums are consistent). Thank you for testing!
Quoted 1 line
This is 2.4.6, on a KT7-RAID board.
So it is KT7-RAID, not KT7A-RAID like mine... Could that have something to do with it...
-- v --