Re: x86, ARM, PARISC, PPC, MIPS and Sparc folks please run this

[Russell King]

On Tue, Sep 02, 2003 at 12:57:31PM +0100, Jamie Lokier wrote:
> You say that "reading from the first mapping _should_ return the
> second write value no matter what", but that there's a bug in the
> write buffer and it isn't doing that.
> 
> I'm saying that the bug can't be that, because such a bug would affect
> normal applications.

I know of no other explaination which fits with the information I have
available to me here.  If you'd care to speculate further, you may,
but I see further speculation as being rather academic, unless it comes
from one of the people who designed the chip.

All this is, however, immateral - the facts are that the write buffer
is buggy, this test detects it, and we can take fairly easy measures
to ensure we fix it up.

Multiple mappings of the same object rarely occur in my experience, so
the resulting performance loss caused by working around the cache and
writebuffer is something we can live with.

-- 
Russell King (rmk@xxxxxxxxxxxxxxxx)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
On Tue, Sep 02, 2003 at 12:57:31PM +0100, Jamie Lokier wrote:
> You say that "reading from the first mapping _should_ return the
> second write value no matter what", but that there's a bug in the
> write buffer and it isn't doing that.
> 
> I'm saying that the bug can't be that, because such a bug would affect
> normal applications.

I know of no other explaination which fits with the information I have
available to me here.  If you'd care to speculate further, you may,
but I see further speculation as being rather academic, unless it comes
from one of the people who designed the chip.

All this is, however, immateral - the facts are that the write buffer
is buggy, this test detects it, and we can take fairly easy measures
to ensure we fix it up.

Multiple mappings of the same object rarely occur in my experience, so
the resulting performance loss caused by working around the cache and
writebuffer is something we can live with.

-- 
Russell King (rmk@xxxxxxxxxxxxxxxx)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/