Re: Linux UIO driver cache problem in PowerPC (fix)

Previous thread: [PATCH] mtd: prevent physmap from causing "request_module: runaway loop modprobe net-pf-1" by Gordon Farquharson on Thursday, March 13, 2008 - 12:13 am. (1 message)

Next thread: hackbench regression since 2.6.25-rc by Zhang, Yanmin on Thursday, March 13, 2008 - 12:46 am. (31 messages)
From: Hans-Jürgen
Date: Thursday, March 13, 2008 - 12:19 am

Am Wed, 12 Mar 2008 15:22:59 -0400

Hi Jean-Samuel,
thanks for your report. Please CC LKML if you find bugs like this. I
also added Grant Likely (PPC Xilinx Virtex maintainer), as it seems to

Hm, we already mark the page with VM_IO. That's sufficient on x86 and
arm. I'm not sure whether this is a bug in PPC memory handling. Why do

It should be OK, but unneccesary on other platforms. I have no
objections, but I fear in 6 months we'll see a patch removing that

I'd like to hear the opinion of people really involved in PPC memory

Thanks,
Hans

--

From: Grant Likely
Date: Thursday, March 13, 2008 - 2:53 pm

Hmmm.  I'm not a memory handling expert so I don't know what's going
on here.  Josh, does this look like anything that would be ppc405
related?




-- 
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.
--

From: Juergen Beisert
Date: Friday, March 14, 2008 - 12:41 am

The UIO also doesn't work on a MPC5200 CPU. With the additional 
pgprot_noncached() call now it works as expected.

Juergen
--

From: Hans J. Koch
Date: Friday, March 14, 2008 - 3:33 am

Am Thu, 13 Mar 2008 08:19:32 +0100

Hi Jean-Samuel,
I investigated a bit and found your patch to be correct. I cleaned it
up a bit and sent it to Greg and LKML. I added you as author and also
took the freedom to add your Signed-off-by.

Juergen, Thomas, Grant, if you can confirm this, please add your
signature as well.

Thanks,
--

From: Leon Woestenberg
Date: Saturday, March 15, 2008 - 7:51 pm

Hello,

Could you elaborate why exactly this is needed, maybe even as a
comment in the patch?

It seemed non-trivial enough to rectify adding a comment.

Thanks,
-- 
Leon
--

From: Hans J. Koch
Date: Sunday, March 16, 2008 - 5:20 am

Am Sun, 16 Mar 2008 03:51:10 +0100

As far as I understood, the VM_IO flag simply means that the page is
used for IO. On some architectures this automatically means that the
page is not cached, but this is not guaranteed. That's why
pgprot_noncached() exists.

If I misunderstood something or somebody can explain it better, please

It _is_ trivial. It's just something you don't need every day, and
therefore it's not known to everybody (including me) or not obvious.
But if you look at it:

VM_IO means IO
pgprot_noncached means noncached

What could be more trivial?

Thanks,
Hans
--

Previous thread: [PATCH] mtd: prevent physmap from causing "request_module: runaway loop modprobe net-pf-1" by Gordon Farquharson on Thursday, March 13, 2008 - 12:13 am. (1 message)

Next thread: hackbench regression since 2.6.25-rc by Zhang, Yanmin on Thursday, March 13, 2008 - 12:46 am. (31 messages)