| View previous topic :: View next topic |
| Author |
Message |
luisfeser
Guru
Joined: 22 May 2004
Posts: 543
Location: /España/Toledo
|
 Posted: Tue Sep 07, 2004 1:40 pm Post subject: Problem with SATA sii3112 in kernel >=2.6.8 |
 |
|
Hello,
I have a SATA seagate 160Gb, and it is configured as scsi (/dev/sda) in a Abit AN7 mobo with sii3112 sata controller.
But with kernel 2.6.9-rc1-mm or nitro, i have this:
| Code: | /dev/sda:
Timing buffer-cache reads: 1660 MB in 2.00 seconds = 829.71 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported
Timing buffered disk reads: 72 MB in 3.07 seconds = 23.44 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported |
And with 2.6.8.1-nitro or ck6, i have this:
| Code: | /dev/sda:
Timing cached reads: 1640 MB in 2.00 seconds = 818.49 MB/sec
Timing buffered disk reads: 70 MB in 3.09 seconds = 22.69 MB/sec |
don't fails, but the result is bad too.
What's wrong???
Anybody with same kernel and driver can put the result of "hdparm -Tt /dev/sdX"??
_________________
AMD Athlon XP-M 2500+ @2200MhHz|1GB DDR @400MHz Dual-Chanel|NVIDIA GeForce4 MX 420|SB Live 5.1|ADSL 512/128 Arsys
gentoo ~x86 |
|
| Back to top |
|
 |
luisfeser
Guru
Joined: 22 May 2004
Posts: 543
Location: /España/Toledo
|
| Posted: Tue Sep 07, 2004 8:34 pm Post subject: |
|
|
Nobody knows?
_________________
AMD Athlon XP-M 2500+ @2200MhHz|1GB DDR @400MHz Dual-Chanel|NVIDIA GeForce4 MX 420|SB Live 5.1|ADSL 512/128 Arsys
gentoo ~x86 |
|
| Back to top |
|
|
luisfeser
Guru
Joined: 22 May 2004
Posts: 543
Location: /España/Toledo
|
| Posted: Wed Sep 08, 2004 11:29 am Post subject: |
|
|
Well, the problem began in 2.6.8 kernel and newer. In the changelog say:
| Quote: | <jgarzik@pobox.com>
[libata sata_sil] Re-fix mod15write bug
Certain early SATA drives have problems with write requests whose
length satisfy the equation "sectors % 15 == 1", on the SiI 3112.
Other drives, and other SiI controllers, are not affected.
The fix for this problem is to avoid such requests, in one of three
ways, for the affect drive+controller combos:
1) Limit all writes to 15 sectors
2) Use block layer features to avoid creating requests whose
length satisfies the above equation.
3) When a request satisfies the above equation, split the request
into two writes, neither of which satisfies the equation.
I chose fix #1, the most simple to implement. After discussion with
Silicon Image and others regarding the impact of this fix, I have
decided to remain with fix #1, and will not be implementing a
"better fix". This means that the affected SATA drives will see
decreased performance, but set of affected drives is small and will
never grow larger.
Further, the complexity of implementing solution #2 or
solution #3 is rather large.
When implementing lba48 'large request' support, I unintentionally
broke the fix for these affected drives. Kudos to Ricky Beam for
noticing this.
This change restores the fix, by adding a flag ATA_DFLAG_LOCK_SECTORS
to indicate that the max_sectors value set by the low-level driver
should never be changed.
|
And i think my seagate is in the blacklist 
Now i am with ide sil drivers:
| Code: | /dev/hde:
Timing cached reads: 1528 MB in 2.00 seconds = 762.97 MB/sec
Timing buffered disk reads: 164 MB in 3.01 seconds = 54.40 MB/sec |
_________________
AMD Athlon XP-M 2500+ @2200MhHz|1GB DDR @400MHz Dual-Chanel|NVIDIA GeForce4 MX 420|SB Live 5.1|ADSL 512/128 Arsys
gentoo ~x86 |
|
| Back to top |
|
|
tgnb
Apprentice
Joined: 16 Apr 2002
Posts: 208
Location: New York, NY
|
| Posted: Sun Dec 05, 2004 8:11 am Post subject: |
|
|
Hmm i think i have this same problem .. hdparm shows terrible performance with kernel's greater than 2.6.7 with my sata / sil / seagate combo.
But i have no HD related problems in 2.6.7 so what does this "fix" do for me other than slow down performance? |
|
| Back to top |
|
|
sir_tez
n00b
Joined: 26 Mar 2003
Posts: 69
|
| Posted: Mon Dec 06, 2004 7:54 am Post subject: |
|
|
| I second that, I too am I using sil3112 sata and that's not an acceptable "solution" for me. |
|
| Back to top |
|
|
sir_tez
n00b
Joined: 26 Mar 2003
Posts: 69
|
| Posted: Tue Dec 07, 2004 4:17 pm Post subject: |
|
|
| Bumping and hoping for a fix. Could more Sil 3112 users post here to keep the thread alive and be counted? |
|
| Back to top |
|
|
sir_tez
n00b
Joined: 26 Mar 2003
Posts: 69
|
| Posted: Tue Dec 07, 2004 6:01 pm Post subject: |
|
|
Say, if other folks who have silicon image 3112 sata controllers could post here and show some solidarity, that'd be awesome  |
|
| Back to top |
|
|
tauruswho
n00b
Joined: 05 Feb 2004
Posts: 29
|
| Posted: Tue Dec 07, 2004 6:12 pm Post subject: |
|
|
Hi All
I have had the same problem for some time with Sil3114 on my lanparty NF2 Ultra, but have used it as is, should I be worried. I did post a while back but was told it was to do with the driver, which seems to be the correct answer. Two sumsung 160G and WD 120G drives all give the same error. Should I revert to to the IDE driver?
Is my data safe from coruption????????????
 |
|
| Back to top |
|
|
sir_tez
n00b
Joined: 26 Mar 2003
Posts: 69
|
| Posted: Wed Dec 08, 2004 10:55 pm Post subject: |
|
|
| Bump |
|
| Back to top |
|
|
Loke
Apprentice
Joined: 25 May 2002
Posts: 274
Location: Norway
|
| Posted: Sat Dec 18, 2004 12:54 am Post subject: |
|
|
This doesnt just apply to the SI3112 chipset and seagate disks, Ive got the SI3512 and Im experiencing the same using seagate disks also. I dont know if the hdparm results can be trusted really, because for every consecutive run of hdparm my "Timing buffered disk read" increases with about 15MB/s.
As another note, my system sure doesnt feel as slow as hdparm say it is.. Actually it feels quite snappy, but I would rather have this fixed of course.
_________________
I'm not saying there should be capital punishment for stupidity, I'm saying why don't we take the warning labels off of everything, and let the problem take care of itself? |
|
| Back to top |
|
|
El_Presidente_Pufferfish
Veteran
Joined: 11 Jul 2002
Posts: 1179
Location: Seattle
|
| Posted: Sat Dec 18, 2004 2:22 am Post subject: |
|
|
my wd 250gb seems affected
| Code: | hdparm -Tt /dev/sda
/dev/sda:
Timing cached reads: 1040 MB in 2.00 seconds = 519.30 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported
Timing buffered disk reads: 146 MB in 3.01 seconds = 48.51 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported |
|
|
| Back to top |
|
|
Loke
Apprentice
Joined: 25 May 2002
Posts: 274
Location: Norway
|
| Posted: Sat Dec 18, 2004 2:33 am Post subject: |
|
|
I just timed the copying of a 1.7GB file from my /home partition to the / (root) partition , using:
| gentoo-dev-source-2.6.9-r9 wrote: |
real 4m55.921s
user 0m0.003s
sys 0m6.576s
|
| SimplyMepis 2.6.7 wrote: |
real 1m33.339s
user 0m0.008s
sys 0m5.259s
|
Both using the SI3512 chip and Seagate ST3160023AS, through the scsi sata_sil interface. The difference is so outstanding, I dont even bother testing anymore. If Im reading the LKML lists correctly, Jeff Garzek the libata guy, thinks this performance loss is acceptable, given the alternative of data corruption. More importantly, this wont get fixed in the future either because it involves quite alot of work!
Im gonna have a hard time talking someone with this chip into using linux, if 3x less HDD performance is what they have to accept. Do the libata guys even know how many MBs are out there with SI chips?
Personally, Im  right now...
_________________
I'm not saying there should be capital punishment for stupidity, I'm saying why don't we take the warning labels off of everything, and let the problem take care of itself? |
|
| Back to top |
|
|
dsd
Developer
Joined: 30 Mar 2003
Posts: 2162
Location: nr London
|
| Posted: Sat Dec 18, 2004 3:01 am Post subject: |
|
|
it is not the sil chips. it is the _combination) of one sil chip with one of a few models of seagate hard disk. every other disk is fine.
those seagate disks arent being produced any more, as far as i know. personally i would try and get the disk swapped.
_________________
http://dev.gentoo.org/~dsd |
|
| Back to top |
|
|
Loke
Apprentice
Joined: 25 May 2002
Posts: 274
Location: Norway
|
| Posted: Sat Dec 18, 2004 3:13 am Post subject: |
|
|
| dsd wrote: | it is not the sil chips. it is the _combination) of one sil chip with one of a few models of seagate hard disk. every other disk is fine.
those seagate disks arent being produced any more, as far as i know. personally i would try and get the disk swapped. |
Yes, I know. Ive been looking at the source for sata_sil.c right now, and it seems there are about 10 seagate disks affected. However, this bug should only apply to SiI3112 chips and those seagate disks blacklisted, according to Jeff Garzik. But the driver just maps the SiI3512 ID and treats it as a SiI3112 chip (which perhaps was just a hack to enable support for the 3512 very quickly..?)
So now Im wondering: Does this bug also affect the SiI3512?
_________________
I'm not saying there should be capital punishment for stupidity, I'm saying why don't we take the warning labels off of everything, and let the problem take care of itself? |
|
| Back to top |
|
|
El_Presidente_Pufferfish
Veteran
Joined: 11 Jul 2002
Posts: 1179
Location: Seattle
|
| Posted: Sat Dec 18, 2004 8:42 pm Post subject: |
|
|
Since the blacklist seems to only affect seagate drives( from looking at the source )
Anybody have a clue why my WD2500JD( 250GB sata ) has the same symptoms? |
|
| Back to top |
|
|
Zyne
Guru
Joined: 08 Jun 2004
Posts: 334
|
| Posted: Sat Dec 18, 2004 10:13 pm Post subject: |
|
|
using 2.6.9-nitro4, having the same issues... 
my hd:
| Code: |
Dec 17 08:32:14 blast Vendor: ATA Model: Maxtor 6Y160M0 Rev: YAR5
Dec 17 08:32:14 blast Type: Direct-Access ANSI SCSI revision: 05
|
I'm gonna try and switch back to 2.6.8.1-love sources...
let's hope it's better...
mobo is Asus A7N8X-deluxe with SIL3112 controller
output of hdparm -Tt /dev/sda
| Code: |
hdparm -Tt /dev/sda
/dev/sda:
Timing cached reads: 1524 MB in 2.00 seconds = 761.35 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported
Timing buffered disk reads: 172 MB in 3.02 seconds = 56.91 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported
|
|
|
| Back to top |
|
|
dsd
Developer
Joined: 30 Mar 2003
Posts: 2162
Location: nr London
|
| Posted: Sat Dec 18, 2004 10:48 pm Post subject: |
|
|
that performance looks good. what is the problem?
_________________
http://dev.gentoo.org/~dsd |
|
| Back to top |
|
|
tauruswho
n00b
Joined: 05 Feb 2004
Posts: 29
|
| Posted: Sat Dec 18, 2004 11:21 pm Post subject: |
|
|
Only seagate????? I use Samsung spin point and a Western Digital, they both give the same errors, see previous post. I am thinking of trying the ide driver, but does this use the same ata-lib???
Lan party ultra B Sil3114.....
Is it possible to patch the kernel with an older SATA driver? |
|
| Back to top |
|
|
Zyne
Guru
Joined: 08 Jun 2004
Posts: 334
|
| Posted: Sat Dec 18, 2004 11:22 pm Post subject: |
|
|
true... I gotta admit those results are looking pretty good...
However, take a look now, as I've done the test 3 times in a row:
| Code: |
# hdparm -tT /dev/sda
/dev/sda:
Timing cached reads: 1020 MB in 2.00 seconds = 509.57 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported
Timing buffered disk reads: 488 MB in 3.00 seconds = 162.42 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported
|
probably a hdparm bug, but iirc hdparm used to work great on my 2.6.8.1-love1 kernel... |
|
| Back to top |
|
|
hw-tph
l33t
Joined: 08 Jan 2004
Posts: 768
Location: Uppsala, Sweden
|
| Posted: Sat Dec 18, 2004 11:42 pm Post subject: |
|
|
My lightly patched vanilla 2.6.9 works great with Sil3112 and a Seagate disk.
Just for sanity checking - do thesee patchsets you all use contain Garzik's CVS/bk libata updates? What about vanilla 2.6.9 - does that work for you?
Håkan |
|
| Back to top |
|
|
nic01
Tux's lil' helper
Joined: 17 Mar 2004
Posts: 87
Location: Copenhagen
|
| Posted: Sun Dec 19, 2004 9:57 am Post subject: |
|
|
Hey
I have the same board and disk as zyne | Quote: | my hd:
Code:
Dec 17 08:32:14 blast Vendor: ATA Model: Maxtor 6Y160M0 Rev: YAR5
Dec 17 08:32:14 blast Type: Direct-Access ANSI SCSI revision: 05
I'm gonna try and switch back to 2.6.8.1-love sources...
let's hope it's better...
mobo is Asus A7N8X-deluxe with SIL3112 controller
output of hdparm -Tt /dev/sda |
except that I run raid 0 and 1 across two disks and 2.6.9-ck3 kernel.
I get the same error when I test the 'raw' disk.
| Code: | /dev/sda:
Timing cached reads: 1612 MB in 2.00 seconds = 804.51 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported
Timing buffered disk reads: 420 MB in 3.03 seconds = 138.54 MB/sec
BLKFLSBUF failed: Operation not supported
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Operation not supported
|
But when I access my md devices it doesn't complain:
| Code: | /dev/md1:
Timing cached reads: 1636 MB in 2.00 seconds = 817.31 MB/sec
Timing buffered disk reads: 252 MB in 3.00 seconds = 83.87 MB/sec
|
Beside that the system runs fine, so I can't complain
/nic |
|
| Back to top |
|
|
Loke
Apprentice
Joined: 25 May 2002
Posts: 274
Location: Norway
|
| Posted: Sun Dec 19, 2004 7:36 pm Post subject: |
|
|
Perhaps Im stating the obvious now, but the problem described in this thread is NOT one of the following:
- Various errors when running hdparm, like blahblahblah: Operation not permitted
- Timing buffered results seems to double for each consecutive run.
The problem in this thread is related to 10 something Seagate HDDs combined with the SiI3112 raid chipset, which result in 10-15 MB/sec buffered disk read due to an errata patch introduced to prevent data corruption.
Unless you have one of these Seagate drives, the driver is working full speed - regardless of what hdparm says! So If you get buffered disk reads in the 45 - 60 MB/sec range, you are already operation at full speed and nothing needs to be done.
Seems quite a few people misunderstood this...
Edit: This means: El_Presidente_Pufferfish, Zyne and a couple of other people posting in this thread - you do NOT have a problem with your setup. The error messages shown in hdparm is a bug with hdparm, and not your driver setup.
_________________
I'm not saying there should be capital punishment for stupidity, I'm saying why don't we take the warning labels off of everything, and let the problem take care of itself? |
|
| Back to top |
|
|
tauruswho
n00b
Joined: 05 Feb 2004
Posts: 29
|
| Posted: Sun Dec 19, 2004 8:38 pm Post subject: |
|
|
Yes But! These errors only started to happen with later kernels, my system was fine (hdparm as well) utill I upgraded to later kernels, I get the same errors as the first post, although my speed seems ok.
This also seems to be the same with Fedora, mandrake etc......but again only on the later kernels. I don't think I have updated hdparm recently.
Anyway do you mean, hdparm only has a bug with later kernels, what should I do? just assume that hdparm is telling me porky pies and hope it is not a serious bug in SATA drivers.
This sort of error message is not what linux needs to sway people to use it, as it has made me question my data safty........ |
|
| Back to top |
|
|
Loke
Apprentice
Joined: 25 May 2002
Posts: 274
Location: Norway
|
| Posted: Sun Dec 19, 2004 8:55 pm Post subject: |
|
|
| tauruswho wrote: | | This sort of error message is not what linux needs to sway people to use it, as it has made me question my data safty........ |
Like I said, the "error" in this thread is transfer speeds of about 10 - 15 MB/sec. If you have any other sane number like 35 - 60 you are ok. Better yet just dmesg|grep errata yourself and see if you get a hit - if yes, the driver is operating in "safemode" and you are affected.
Everything else related to hdparm output - just ignore for now.
_________________
I'm not saying there should be capital punishment for stupidity, I'm saying why don't we take the warning labels off of everything, and let the problem take care of itself? |
|
| Back to top |
|
|
luisfeser
Guru
Joined: 22 May 2004
Posts: 543
Location: /España/Toledo
|
| Posted: Sun Jan 09, 2005 12:32 pm Post subject: |
|
|
Hello all
The problem with silicon 3112 and my seagate is worth with 2.6.10
Before, with IDE driver of silicon was working well, but now i get a "Timing buffered disk reads" around 20MB/sec
With scsi driver i get 22MB/sec, still a bad result.
maybe some tweak for hdparm for increase performance???
_________________
AMD Athlon XP-M 2500+ @2200MhHz|1GB DDR @400MHz Dual-Chanel|NVIDIA GeForce4 MX 420|SB Live 5.1|ADSL 512/128 Arsys
gentoo ~x86 |
|
| Back to top |
|
|
|
Kernel & Hardware
