sis 5591 ide in 2.4.19-pre3 consumes souls
yesterday night i excitedly came home to my shiny new -pre3, to discover that on boot, all ide devices timeout on DMA requests. i'm using the sis ide driver on my sis 5591 controller, which is integrated onto an amptron 810 motherboard.
i haven't had a chance to capture exact logs yet, but can do testing tonight. exact behavior follows:
rebooted with 2.4.19-pre3-jl11 (preemptive; jl11 is a patch to networking code that should play no role here). hard drive hda and cdrom's hdc,hdd are detected. upon partition check of hda1, the system hung for roughly 20 seconds, after which it was declared dma commands had timed out. this was repeated for all other partitions on the drive, after which the kernel panicked, unable to mount the root fs.
this behavior continued to manifest over any number of reboots. interestingly, i could get my old 2.4.18-jl11 (non-preempt) to work fine, but only after a hard power down. reboots left it in the same situation, but not logging nearly so much to the console :).
Signature · 3 lines
nicholas black ([email protected]) "c has types for a reason. c++ improved the type system for a reason. perl and php programs have run-time failures for a reason." - lkml
Quoted 5 lines · nicholas black
yesterday night i excitedly came home to my shiny new -pre3, to discover that on boot, all ide devices timeout on DMA requests. i'm using the sis ide driver on my sis 5591 controller, which is integrated onto an amptron 810 motherboard.
There's definitely a problem with some UDMA 33 chipsets. I'll ask SiS for 5591 chipset info RSN and look into this as soon as I find some time.
Quoted 11 lines
i haven't had a chance to capture exact logs yet, but can do testing tonight. exact behavior follows:
rebooted with 2.4.19-pre3-jl11 (preemptive; jl11 is a patch to networking code that should play no role here). hard drive hda and cdrom's hdc,hdd are detected. upon partition check of hda1, the system hung for roughly 20 seconds, after which it was declared dma commands had timed out. this was repeated for all other partitions on the drive, after which the kernel panicked, unable to mount the root fs.
Quoted 5 lines
this behavior continued to manifest over any number of reboots. interestingly, i could get my old 2.4.18-jl11 (non-preempt) to work fine, but only after a hard power down. reboots left it in the same situation, but not logging nearly so much to the console :).
Could you send us the output of "lspci -vxxx" with 2.4.18-jl11 and boot logs of 2.4.19-pre3-jl11 please ?
Did you "make oldconfig" on the 2.4.19-pre3-jl11 sources with the your old 2.4.18-jl11 .config or made a new config from scratch ? If so assuming you didn't modify your source trees could you send us the result of :
diff <2.4.18-jl11_tree>/.config <2.4.19-pre3-jl11_tree>/.config | grep "^. CONFIG"
LB.
Quoted 2 lines · In article <[email protected]> you wrote:
There's definitely a problem with some UDMA 33 chipsets. I'll ask SiS for 5591 chipset info RSN and look into this as soon as I find some time.
i look forward to it :). in light of this problem (i've recieved mail from one michal dorocinski echoing this behavior and referencing an -ac patch; i've asked him to mail it here), will this driver stay in .19 for the time being?
Quoted 2 lines
Could you send us the output of "lspci -vxxx" with 2.4.18-jl11 and boot logs of 2.4.19-pre3-jl11 please ?
i shall do as much tonight, when i'm at the machine.
Quoted 2 lines
Did you "make oldconfig" on the 2.4.19-pre3-jl11 sources with the your old 2.4.18-jl11 .config or made a new config from scratch ?
of course; i don't have time for the latter :). i just recreated this, and here were applicable questions from oldconfig, and my responses:
no to auto-geometry resizing support
no to ide taskfile access
no to force enable legacy hosts to dma
no to dma only
no to noop elevator
the previous setup enabled generic pci ide chipsets, sharing of pci ide interrupts, generic pci bus-master, use dma by default, and sis5513. nothing else regarding ide chipset support/bugfixes is enabled.
Quoted 4 lines
If so assuming you didn't modify your source trees could you send us the result of :
diff <2.4.18-jl11_tree>/.config <2.4.19-pre3-jl11_tree>/.config | grep "^. CONFIG"
using said simulated .configs:
[ouzo](1) $ diff linux-2.4.18/.config linux-2.4.19-pre3/.config | grep "^. CONFIG"
Quoted 1 line
CONFIG_X86_MCE=y
< CONFIG_PACKET_MMAP_FILTER=y
< CONFIG_BLK_DEV_ADMA=y
Quoted 1 line
CONFIG_BLK_DEV_ADMA=y
[ouzo](0) $
one other perhaps relevant piece of information: this is not the finest hard drive in the world; some months ago i broke off the grounding 40 pin. that line of the ide cable is now spliced off, and wrapped around a screw :).
Signature · 3 lines
nicholas black ([email protected]) "c has types for a reason. c++ improved the type system for a reason. perl and php programs have run-time failures for a reason." - lkml
Quoted 8 lines
Quoted 2 lines · In article <[email protected]> you wrote:
There's definitely a problem with some UDMA 33 chipsets. I'll ask SiS for 5591 chipset info RSN and look into this as soon as I find some time.
i look forward to it :). in light of this problem (i've recieved mail from one michal dorocinski echoing this behavior and referencing an -ac patch; i've asked him to mail it here), will this driver stay in .19 for the time being?
Yes.
As I found it in -pre3 the result was: timeouts and some debugs (Lionel has the log) and no partition table was found (BadCRC messages from hda). It worked then after a day or two when -ac1 or -ac2 was found so I think it was solved. I will check -pre4 if it work.
Btw. I have a question... will an SiS 6326 AGP will be supported anyhow by agp gart ? It is supported by Xfree driver but no xv nor AGP or dri acceleration as I can see is realy used. It also has some wired problems with colorded dots (blue and red).
But this is maybe a subject to AGP/SiS VGA/XFREE drivers maintaners :)
Greetings
Zwierz
Signature · 2 lines
The Shadow, The Darkness, The Fear... Forever Alone Immortal...
Hello,
As I wrote I tryed 2.4.19-pre5 on my SIS motherboard nad 5513 IDE chipset and again I got the same problems. The -ac series of 2.4.19-pre4 is good and works fine.
Greetings
Michal 'Zwerz' Dorocinski
Signature · 2 lines
The Shadow, The Darkness, The Fear... Forever Alone Immortal...
Quoted 6 lines · Michal Dorocinski
Hello,
As I wrote I tryed 2.4.19-pre5 on my SIS motherboard nad 5513 IDE chipset and again I got the same problems. The -ac series of 2.4.19-pre4 is good and works fine.
Marcelo, Alan, do you want me to provide a patch against pre5 or is the latest sis5513 patch in AC's tree already scheduled for inclusion in Marcelo's tree ?
Reminder: 2.4.19-pre5's sis5513.c holds 2 show-stoppers :
- incorrect SiS730 support,
- for most chipsets : lack of proper UDMA init if the BIOS doesn't setup UDMA itself.
The latest is the one bitting Michal IIRC.
LB.