I have an old ISA-PNP soundcard whose driver doesn't currently use the
ISA-PNP services. I have therefore been activating this card by
writing to the /proc/isapnp interface. And in an idle moment, I tried
an "auto" instruction:
cat > /proc/isapnp << EOF
card 0 ENS3081
dev 0 ENS0000
auto
EOF
This operation didn't return and system CPU usage then climbed to
100%, although the box didn't lock up. Further investigations showed
that the problem happened when either the isapnp_config_activate() or
isapnp_config_deactivate() functions tried to acquire the
isapnp_cfg_mutex. This seemed to be because they tried to acquire the
mutex for the specific devfn (zero in this case), but the
isapnp_set_card() function has already acquired it using -1 instead.
Couldn't the isapnp_autoconfigure() function use isapnp_activate() and
isapnp_deactivate() instead of isapnp_info_device->activate() and
isapnp_info_device->deactivate()? The former don't try to acquire the
mutex, and are precisely what the "activate" and "deactivate" /proc
commands invoke anyway.
Cheers,
Chris
This is the output from /proc/isapnp:
Card 1 'ENS3081:ENSONIQ Soundscape' PnP version 1.0
Logical device 0 'ENS0000:Unknown'
Device is not active
Active port 0x330,0x300
Active IRQ 5 [0x2],9 [0x2]
Active DMA 1,3
Resources 0
Priority preferred
Port 0x330-0x330, align 0xf, size 0x10, 16-bit address decoding
IRQ 5,7 High-Edge
IRQ 2/9 High-Edge
DMA 1 8-bit byte-count compatible
DMA 0,3 8-bit byte-count compatible
Alternate resources 0:1
Priority acceptable
Port 0x330-0x330, align 0xf, size 0x10, 16-bit address decoding
IRQ 5,7,2/9 High-Edge
IRQ 7,2/9,10 High-Edge
DMA 1,3 8-bit byte-count compatible
DMA 0,3 8-bit byte-count compatible
Alternate resources 0:2
Priority functional
Port 0x330-0x330, align 0xf, size 0x10, 16-bit address decoding
IRQ 5,7,2/9 High-Edge
IRQ 7,2/9,10 High-Edge
DMA 0,1,3 8-bit byte-count compatible
Alternate resources 0:3
Priority functional
Port 0x350-0x350, align 0xf, size 0x10, 16-bit address decoding
IRQ 5,7,2/9 High-Edge
IRQ 7,2/9,10 High-Edge
DMA 0,1,3 8-bit byte-count compatible
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/