regression between 2.4.18 and 2.4.21/22

regression between 2.4.18 and 2.4.21/22

[Justin Hibbits]

Something very strange is going on with my machine.  With 2.4.18, I was 
getting 38MB/s on my main system disk (IBM Deskstar 60gxp), and 35 for 
the other drives (Western Digital).  The IBM drive is on a Promise IDE 
controller (ASUS A7V266-E motherboard), and the others are on a PROMISE 
2069 UDMA133 controller.  However, with 2.4.21 and 2.4.22, it will not 
set the using_dma flag for my IBM drive, but sets it for the others, 
which now get sustained transfer rates of 46MB/s or greater.  I'm using 
the same options for all 3 kernels (at least, for the ATA/IDE options). 
  Any help would be appreciated, and I'll see if maybe I could do 
something with it when I get time.

Thanks,

Justin

Re: regression between 2.4.18 and 2.4.21/22

[Mike Fedyk]

On Sun, Oct 05, 2003 at 04:21:06PM -0400, Justin Hibbits wrote:
> which now get sustained transfer rates of 46MB/s or greater.  I'm using 
> the same options for all 3 kernels (at least, for the ATA/IDE options). 
>  Any help would be appreciated, and I'll see if maybe I could do 
> something with it when I get time.

Some drivers have been split, or renamed.  Make sure you have the driver for
your chipset compiled in, and you're not using generic ide dma.

Re: regression between 2.4.18 and 2.4.21/22

[Justin Hibbits]

On Sunday, Oct 5, 2003, at 17:58 America/New_York, Mike Fedyk wrote:

> On Sun, Oct 05, 2003 at 04:21:06PM -0400, Justin Hibbits wrote:
>> which now get sustained transfer rates of 46MB/s or greater.  I'm 
>> using
>> the same options for all 3 kernels (at least, for the ATA/IDE 
>> options).
>>  Any help would be appreciated, and I'll see if maybe I could do
>> something with it when I get time.
>
> Some drivers have been split, or renamed.  Make sure you have the 
> driver for
> your chipset compiled in, and you're not using generic ide dma.

I'm using the PROMISE drivers (the only ones listed in the kernel 
config), and Generic is also compiled in.  It wasn't compiled in 
2.4.21, so I tried compiling it in, and it still didn't work, tried 
changing the "Ignore word93 Validation Bits" option, but that still 
didn't help.  Guess I should just take a crack at the source itself.

-Justin

Re: regression between 2.4.18 and 2.4.21/22

[Bartlomiej Zolnierkiewicz]

Please narrow down kernel version if you want your problem to be cared.

Try 2.4.19, 2.4.20.  There are also intermediate prepatches at
http://www.kernel.org/pub/linux/kernel/v2.4/testing/old/

dmesg output and .config can also be useful.

--bartlomiej

On Sunday 05 of October 2003 22:21, Justin Hibbits wrote:
> Something very strange is going on with my machine.  With 2.4.18, I was
> getting 38MB/s on my main system disk (IBM Deskstar 60gxp), and 35 for
> the other drives (Western Digital).  The IBM drive is on a Promise IDE
> controller (ASUS A7V266-E motherboard), and the others are on a PROMISE
> 2069 UDMA133 controller.  However, with 2.4.21 and 2.4.22, it will not
> set the using_dma flag for my IBM drive, but sets it for the others,
> which now get sustained transfer rates of 46MB/s or greater.  I'm using
> the same options for all 3 kernels (at least, for the ATA/IDE options).
>   Any help would be appreciated, and I'll see if maybe I could do
> something with it when I get time.

Re: regression between 2.4.18 and 2.4.21/22

[Justin Hibbits]

[-- Attachment #1: Type: text/plain, Size: 1584 bytes --]


On Sunday, Oct 5, 2003, at 19:22 America/New_York, Bartlomiej 
Zolnierkiewicz wrote:

>
> Please narrow down kernel version if you want your problem to be cared.
>
> Try 2.4.19, 2.4.20.  There are also intermediate prepatches at
> http://www.kernel.org/pub/linux/kernel/v2.4/testing/old/
>
> dmesg output and .config can also be useful.
>
> --bartlomiej
>
> On Sunday 05 of October 2003 22:21, Justin Hibbits wrote:
>> Something very strange is going on with my machine.  With 2.4.18, I 
>> was
>> getting 38MB/s on my main system disk (IBM Deskstar 60gxp), and 35 for
>> the other drives (Western Digital).  The IBM drive is on a Promise IDE
>> controller (ASUS A7V266-E motherboard), and the others are on a 
>> PROMISE
>> 2069 UDMA133 controller.  However, with 2.4.21 and 2.4.22, it will not
>> set the using_dma flag for my IBM drive, but sets it for the others,
>> which now get sustained transfer rates of 46MB/s or greater.  I'm 
>> using
>> the same options for all 3 kernels (at least, for the ATA/IDE 
>> options).
>>   Any help would be appreciated, and I'll see if maybe I could do
>> something with it when I get time.

Ok, I tried 2.4.19, which I thought was pretty bad because it randomly 
crashed all the time, and it worked just fine with all my drives.  
2.4.20 with the wolk-4.0 patch also worked.  So, I'm guessing it was 
between 2.4.20 and 2.4.21....I could try all the prepatches as well, 
and narrow down exact prepatch, will take some time.  dmesg output for 
2.4.21 follows (uses a patchset for XFS, sensors, etc), along with my 
config, both compressed.


[-- Attachment #2: .config.bz2 --]
[-- Type: application/octet-stream, Size: 5193 bytes --]

[-- Attachment #3: dmesg.log.bz2 --]
[-- Type: application/octet-stream, Size: 5803 bytes --]

[-- Attachment #4: Type: text/plain, Size: 1 bytes --]

Re: regression between 2.4.18 and 2.4.21/22

[Bartlomiej Zolnierkiewicz]

Your /dev/hda (IBM DeskStar 60GXP) is not in DMA mode because
you don't have support for your IDE controller compiled-in.
Going from 2.4.21 you have to explicitely enable support for IDE chipsets.
Assumption that current .config file will work with future kernel versions
is not true.  Please compile kernel with driver for your on-board IDE chipset
(I deducted from your dmesg that it is VIA82CXXX IDE driver).

Please report back if this cures your problem,

Thanks,
--bartlomiej

On Monday 06 of October 2003 01:38, Justin Hibbits wrote:
> On Sunday, Oct 5, 2003, at 19:22 America/New_York, Bartlomiej
>
> Zolnierkiewicz wrote:
> > Please narrow down kernel version if you want your problem to be cared.
> >
> > Try 2.4.19, 2.4.20.  There are also intermediate prepatches at
> > http://www.kernel.org/pub/linux/kernel/v2.4/testing/old/
> >
> > dmesg output and .config can also be useful.
> >
> > --bartlomiej
> >
> > On Sunday 05 of October 2003 22:21, Justin Hibbits wrote:
> >> Something very strange is going on with my machine.  With 2.4.18, I
> >> was
> >> getting 38MB/s on my main system disk (IBM Deskstar 60gxp), and 35 for
> >> the other drives (Western Digital).  The IBM drive is on a Promise IDE
> >> controller (ASUS A7V266-E motherboard), and the others are on a
> >> PROMISE
> >> 2069 UDMA133 controller.  However, with 2.4.21 and 2.4.22, it will not
> >> set the using_dma flag for my IBM drive, but sets it for the others,
> >> which now get sustained transfer rates of 46MB/s or greater.  I'm
> >> using
> >> the same options for all 3 kernels (at least, for the ATA/IDE
> >> options).
> >>   Any help would be appreciated, and I'll see if maybe I could do
> >> something with it when I get time.
>
> Ok, I tried 2.4.19, which I thought was pretty bad because it randomly
> crashed all the time, and it worked just fine with all my drives.
> 2.4.20 with the wolk-4.0 patch also worked.  So, I'm guessing it was
> between 2.4.20 and 2.4.21....I could try all the prepatches as well,
> and narrow down exact prepatch, will take some time.  dmesg output for
> 2.4.21 follows (uses a patchset for XFS, sensors, etc), along with my
> config, both compressed.

Re: regression between 2.4.18 and 2.4.21/22

[Justin Hibbits]

On Monday, Oct 6, 2003, at 15:06 America/New_York, Bartlomiej 
Zolnierkiewicz wrote:

>
> Your /dev/hda (IBM DeskStar 60GXP) is not in DMA mode because
> you don't have support for your IDE controller compiled-in.
> Going from 2.4.21 you have to explicitely enable support for IDE 
> chipsets.
> Assumption that current .config file will work with future kernel 
> versions
> is not true.  Please compile kernel with driver for your on-board IDE 
> chipset
> (I deducted from your dmesg that it is VIA82CXXX IDE driver).
>
> Please report back if this cures your problem,
>
> Thanks,
> --bartlomiej
>
> On Monday 06 of October 2003 01:38, Justin Hibbits wrote:
>> On Sunday, Oct 5, 2003, at 19:22 America/New_York, Bartlomiej
>>
>> Zolnierkiewicz wrote:
>>> Please narrow down kernel version if you want your problem to be 
>>> cared.
>>>
>>> Try 2.4.19, 2.4.20.  There are also intermediate prepatches at
>>> http://www.kernel.org/pub/linux/kernel/v2.4/testing/old/
>>>
>>> dmesg output and .config can also be useful.
>>>
>>> --bartlomiej
>>>
>>> On Sunday 05 of October 2003 22:21, Justin Hibbits wrote:
>>>> Something very strange is going on with my machine.  With 2.4.18, I
>>>> was
>>>> getting 38MB/s on my main system disk (IBM Deskstar 60gxp), and 35 
>>>> for
>>>> the other drives (Western Digital).  The IBM drive is on a Promise 
>>>> IDE
>>>> controller (ASUS A7V266-E motherboard), and the others are on a
>>>> PROMISE
>>>> 2069 UDMA133 controller.  However, with 2.4.21 and 2.4.22, it will 
>>>> not
>>>> set the using_dma flag for my IBM drive, but sets it for the others,
>>>> which now get sustained transfer rates of 46MB/s or greater.  I'm
>>>> using
>>>> the same options for all 3 kernels (at least, for the ATA/IDE
>>>> options).
>>>>   Any help would be appreciated, and I'll see if maybe I could do
>>>> something with it when I get time.
>>
>> Ok, I tried 2.4.19, which I thought was pretty bad because it randomly
>> crashed all the time, and it worked just fine with all my drives.
>> 2.4.20 with the wolk-4.0 patch also worked.  So, I'm guessing it was
>> between 2.4.20 and 2.4.21....I could try all the prepatches as well,
>> and narrow down exact prepatch, will take some time.  dmesg output for
>> 2.4.21 follows (uses a patchset for XFS, sensors, etc), along with my
>> config, both compressed.

Thanks Bartlomeij, building that driver helped.  Curious though, since 
it's using the PROMISE chip, or should be, since, according to my 
motherboard's manual, the PROMISE chip is the only IDE chip there.

Re: regression between 2.4.18 and 2.4.21/22

[Bartlomiej Zolnierkiewicz]

On Monday 06 of October 2003 22:21, Justin Hibbits wrote:
> On Monday, Oct 6, 2003, at 15:06 America/New_York, Bartlomiej
>
> Zolnierkiewicz wrote:
> > Your /dev/hda (IBM DeskStar 60GXP) is not in DMA mode because
> > you don't have support for your IDE controller compiled-in.
> > Going from 2.4.21 you have to explicitely enable support for IDE
> > chipsets.
> > Assumption that current .config file will work with future kernel
> > versions
> > is not true.  Please compile kernel with driver for your on-board IDE
> > chipset
> > (I deducted from your dmesg that it is VIA82CXXX IDE driver).
> >
> > Please report back if this cures your problem,
> >
> > Thanks,
> > --bartlomiej
> >
> > On Monday 06 of October 2003 01:38, Justin Hibbits wrote:
> >> On Sunday, Oct 5, 2003, at 19:22 America/New_York, Bartlomiej
> >>
> >> Zolnierkiewicz wrote:
> >>> Please narrow down kernel version if you want your problem to be
> >>> cared.
> >>>
> >>> Try 2.4.19, 2.4.20.  There are also intermediate prepatches at
> >>> http://www.kernel.org/pub/linux/kernel/v2.4/testing/old/
> >>>
> >>> dmesg output and .config can also be useful.
> >>>
> >>> --bartlomiej
> >>>
> >>> On Sunday 05 of October 2003 22:21, Justin Hibbits wrote:
> >>>> Something very strange is going on with my machine.  With 2.4.18, I
> >>>> was
> >>>> getting 38MB/s on my main system disk (IBM Deskstar 60gxp), and 35
> >>>> for
> >>>> the other drives (Western Digital).  The IBM drive is on a Promise
> >>>> IDE
> >>>> controller (ASUS A7V266-E motherboard), and the others are on a
> >>>> PROMISE
> >>>> 2069 UDMA133 controller.  However, with 2.4.21 and 2.4.22, it will
> >>>> not
> >>>> set the using_dma flag for my IBM drive, but sets it for the others,
> >>>> which now get sustained transfer rates of 46MB/s or greater.  I'm
> >>>> using
> >>>> the same options for all 3 kernels (at least, for the ATA/IDE
> >>>> options).
> >>>>   Any help would be appreciated, and I'll see if maybe I could do
> >>>> something with it when I get time.
> >>
> >> Ok, I tried 2.4.19, which I thought was pretty bad because it randomly
> >> crashed all the time, and it worked just fine with all my drives.
> >> 2.4.20 with the wolk-4.0 patch also worked.  So, I'm guessing it was
> >> between 2.4.20 and 2.4.21....I could try all the prepatches as well,
> >> and narrow down exact prepatch, will take some time.  dmesg output for
> >> 2.4.21 follows (uses a patchset for XFS, sensors, etc), along with my
> >> config, both compressed.
>
> Thanks Bartlomeij, building that driver helped.  Curious though, since
> it's using the PROMISE chip, or should be, since, according to my
> motherboard's manual, the PROMISE chip is the only IDE chip there.

Good.  It has also on-board IDE ports (from VIA chipset).

--bartlomiej