2002-07-01 07:20:50

by Samuel Thibault

Subject: [PATCH] 2.4.18 linux/drivers/maestro.c dev_audio flaw

Hi !

It seems that the value of ess->dev_audio is wrongly interpreted. Here's
its only initialization :
(maestro.c:3473)

if ((s->dev_audio = register_sound_dsp(&ess_audio_fops, -1)) < 0)

register_sound_dsp can return -ENOMEM for instance, so checking whether < 0
is mandatory. However, in (almost) the rest of the file, only == -1 is
tested, so here's a patch.

Furthermore, printing some error message may sometimes be handy, such as when
-ENOENT is returned (no hole was found for the driver), just to understand why
the device was probed, but isn't actually accessible.

Best regards,

Samuel Thibault

diff -urN linux-2.4.18-cor/drivers/sound/maestro.c linux-2.4.18-cor2/drivers/sound/maestro.c
--- linux-2.4.18-cor/drivers/sound/maestro.c Sun Jun 30 22:02:11 2002
+++ linux-2.4.18-cor2/drivers/sound/maestro.c Sun Jun 30 23:34:26 2002
@@ -1963,7 +1963,7 @@
for(i=0;i<NR_DSPS;i++)
{
s=&c->channels[i];
- if(s->dev_audio == -1)
+ if(s->dev_audio < 0)
break;
spin_lock(&s->lock);
ess_update_ptr(s);
@@ -2917,7 +2917,7 @@
for(i=0;i<NR_DSPS;i++) {
struct ess_state *ess = &s->card->channels[i];

- if(ess->dev_audio == -1)
+ if(ess->dev_audio < 0)
continue;

ess->dma_dac.ready = s->dma_dac.mapped = 0;
@@ -3480,7 +3480,7 @@
for(;i<NR_DSPS;i++)
{
struct ess_state *s=&card->channels[i];
- s->dev_audio = -1;
+ s->dev_audio = -ENODEV;
}

ess = &card->channels[0];
@@ -3541,7 +3541,7 @@
for(i=0;i<NR_DSPS;i++)
{
struct ess_state *s = &card->channels[i];
- if(s->dev_audio != -1)
+ if(s->dev_audio >= 0)
unregister_sound_dsp(s->dev_audio);
}
release_region(card->iobase, 256);
@@ -3586,7 +3586,7 @@
for(i=0;i<NR_DSPS;i++)
{
struct ess_state *ess = &card->channels[i];
- if(ess->dev_audio != -1)
+ if(ess->dev_audio >= 0)
unregister_sound_dsp(ess->dev_audio);
}
/* Goodbye, Mr. Bond. */
@@ -3690,7 +3690,7 @@
for(i=0;i<NR_DSPS;i++) {
struct ess_state *s = &card->channels[i];

- if(s->dev_audio == -1)
+ if(s->dev_audio < 0)
continue;

M_printk("maestro: stopping apus for device %d\n",i);
@@ -3743,7 +3743,7 @@
struct ess_state *s = &card->channels[i];
int chan,reg;

- if(s->dev_audio == -1)
+ if(s->dev_audio < 0)
continue;

for(chan = 0 ; chan < 6 ; chan++) {

2002-07-01 16:05:15

by Zach Brown

Subject: Re: [PATCH] 2.4.18 linux/drivers/maestro.c dev_audio flaw

> It seems that the value of ess->dev_audio is wrongly interpreted.

Thanks for attacking this driver, Samuel. Your patches have looked good
so far. I'm sorry I'm not more involved; its been quite a while since
I've futzed about with the maestro(s).

You may want to cc: your fixes to Takashi Iwai <[email protected]> , who
actively maintains the ALSA equivalent drivers. He'll have a much
greater chance of affecting change than I do :)

- z

2002-07-01 16:13:10

by Jeff Garzik

Subject: Re: [PATCH] 2.4.18 linux/drivers/maestro.c dev_audio flaw

Zach Brown wrote:
>>It seems that the value of ess->dev_audio is wrongly interpreted.
>
>
> Thanks for attacking this driver, Samuel. Your patches have looked good
> so far. I'm sorry I'm not more involved; its been quite a while since
> I've futzed about with the maestro(s).



I'm afraid I missed the original patches :(

But I am willing to review and apply drivers/sound [sound/oss in 2.5]
patches, if no one else (Alan?) does so.

Regards,

Jeff