Skip to contentSkip to conversation

lx6464es fails to open a second time

bisected Tim Blechmann 3 participants 4 messages Nov 22, 2011

lore
Share & tools
lore
#0 lore raw

hi all, especially brian and takashi,

some time ago, i've been developing the driver for the lx6464es ethersound sound card, which has been included into the kernel for some time. however i haven't been able to use it in kernels after 2.6.33.

today, i was able to bisect the issue and the first bad commit is:

commit 6175ddf06b6172046a329e3abfd9c901a43efd2e

Author: Brian Gerst <[email protected]>
Date:   Fri Feb 5 09:37:07 2010 -0500

x86: Clean up mem*io functions.

Iomem has no special significance on x86. Use the standard mem* functions instead of trying to call other versions. Some fixups are needed to match the function prototypes.

LKML-Reference: <[email protected]>

the problem was: the sound card could only be started once, but i haven't been able to start it a second time, probably because the driver doesn't shut down the device correctly.

the communication with the device is done by passing simple commands via memcpy_fromio and memcpy_toio (compare sound/pci/lx6464es/lx_core.c, lines 75 to 99). any idea, what is going wrong there?

thanks, tim

#1 lore raw
Quoted 7 lines · Tim Blechmann

some time ago, i've been developing the driver for the lx6464es ethersound sound card, which has been included into the kernel for some time. however i haven't been able to use it in kernels after 2.6.33.

today, i was able to bisect the issue and the first bad commit is:

commit 6175ddf06b6172046a329e3abfd9c901a43efd2e

Quoted 8 lines
Author: Brian Gerst <[email protected]>
Date:   Fri Feb 5 09:37:07 2010 -0500

x86: Clean up mem*io functions.

the communication with the device is done by passing simple commands via memcpy_fromio and memcpy_toio (compare sound/pci/lx6464es/lx_core.c, lines 75 to 99). any idea, what is going wrong there?

The old and new implementations of memcpy_*io do not have any differences in the documented API, but the new ones are much more optimized, so they might not use 8- or 32-bit accesses or a different access pattern.

Does the card's mapped I/O region behave exactly like memory, or has it any restrictions on how it can be used? In the latter case, you should implement your readbuf/writebuf functions by hand to use plain 32-bit accesses in order (or whatever is required).

Regards,
Clemens

#2 lore raw
Quoted 12 lines
Quoted 11 lines

today, i was able to bisect the issue and the first bad commit is:

commit 6175ddf06b6172046a329e3abfd9c901a43efd2e

Author: Brian Gerst <[email protected]>
Date:   Fri Feb 5 09:37:07 2010 -0500

x86: Clean up mem*io functions.

the communication with the device is done by passing simple commands via memcpy_fromio and memcpy_toio (compare sound/pci/lx6464es/lx_core.c, lines 75 to 99). any idea, what is going wrong there?

Quoted 8 lines

The old and new implementations of memcpy_*io do not have any differences in the documented API, but the new ones are much more optimized, so they might not use 8- or 32-bit accesses or a different access pattern.

Does the card's mapped I/O region behave exactly like memory, or has it any restrictions on how it can be used? In the latter case, you should implement your readbuf/writebuf functions by hand to use plain 32-bit accesses in order (or whatever is required).

indeed, emulating memcpy_*io with ioread32/iowrite32 fixes the issue.

thanks, tim

#3 lore raw

At Tue, 22 Nov 2011 11:17:45 +0100,

Quoted 17 lines · Tim Blechmann
Quoted 17 lines
Quoted 4 lines

today, i was able to bisect the issue and the first bad commit is:

commit 6175ddf06b6172046a329e3abfd9c901a43efd2e

Quoted 8 lines
Author: Brian Gerst <[email protected]>
Date:   Fri Feb 5 09:37:07 2010 -0500

x86: Clean up mem*io functions.

the communication with the device is done by passing simple commands via memcpy_fromio and memcpy_toio (compare sound/pci/lx6464es/lx_core.c, lines 75 to 99). any idea, what is going wrong there?

The old and new implementations of memcpy_*io do not have any differences in the documented API, but the new ones are much more optimized, so they might not use 8- or 32-bit accesses or a different access pattern.

Quoted 6 lines
Quoted 4 lines

Does the card's mapped I/O region behave exactly like memory, or has it any restrictions on how it can be used? In the latter case, you should implement your readbuf/writebuf functions by hand to use plain 32-bit accesses in order (or whatever is required).

indeed, emulating memcpy_*io with ioread32/iowrite32 fixes the issue.

Good to hear. Could you submit a patch?

thanks,

Takashi