Skip to contentSkip to conversation

ALSA: core: Use the allocated major number with major=0

PATCH In review Ali Ahmet Memiş 3 participants 5 messages Oct 11 – Oct 14 Sound

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ALSA: core: Use the allocated major number with major=0

Since commit bd8f21876189 ("ALSA: core: propagate actual error code from register_chrdev"), alsa_sound_init() only treats negative values from register_chrdev() as errors. With the major=0 module option register_chrdev() allocates a major number dynamically and returns it but snd_major and the devices still use 0. The devices are then created without device numbers so no /dev/snd nodes appear and sound is unusable.

Use the returned major number when major=0 is given.

Fixes bd8f21876189 ("ALSA: core: propagate actual error code from register_chrdev")
---
 sound/core/sound.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
sound/core/sound.c +3 -1
index 24f40763a20c..a0349a7e9495 100644
@@ -396,7 +396,6 @@ static int __init alsa_sound_init(void)
 {
 	int err;
 
-	snd_major = major;
 	snd_ecards_limit = cards_limit;
 
 	err = register_chrdev(major, "alsa", &snd_fops);
@@ -404,6 +403,9 @@ static int __init alsa_sound_init(void)
 		pr_err("ALSA core: unable to register native major device number %d\n", major);
 		return err;
 	}
+	if (!major)
+		major = err;
+	snd_major = major;
 
 	err = snd_info_init();
 	if (err < 0) {
Signature · 1 line
2.55.0
#1 lore raw

Hi Ali , Takashi-san,

Quoted 8 lines

Since commit bd8f21876189 ("ALSA: core: propagate actual error code from register_chrdev"), alsa_sound_init() only treats negative values from register_chrdev() as errors. With the major=0 module option register_chrdev() allocates a major number dynamically and returns it but snd_major and the devices still use 0. The devices are then created without device numbers so no /dev/snd nodes appear and sound is unusable.

Note that major=0 has never worked, even before bd8f21876189: the old code treated the positive return value of register_chrdev() as an error and failed with -EIO. bd8f21876189 only changed the symptom.

So this patch adds support for a dynamic major rather than fixing a regression. Takashi-san, is a dynamic major something ALSA wants to support? If not, rejecting major=0 with -EINVAL might be simpler.

If it is kept, the Fixes tag is fine since the patch depends on bd8f21876189, but please mention in the changelog that major=0 never worked before.

Best regards,
Phuc

#2 lore raw
Quoted 13 lines · Bui Duc Phuc

Hi Ali , Takashi-san,

Quoted 8 lines

Since commit bd8f21876189 ("ALSA: core: propagate actual error code from register_chrdev"), alsa_sound_init() only treats negative values from register_chrdev() as errors. With the major=0 module option register_chrdev() allocates a major number dynamically and returns it but snd_major and the devices still use 0. The devices are then created without device numbers so no /dev/snd nodes appear and sound is unusable.

Quoted 7 lines

Note that major=0 has never worked, even before bd8f21876189: the old code treated the positive return value of register_chrdev() as an error and failed with -EIO. bd8f21876189 only changed the symptom.

So this patch adds support for a dynamic major rather than fixing a regression. Takashi-san, is a dynamic major something ALSA wants to support?

No, it's supposed to be only the fixed one.

If there is a real demand, we can extend the implementation, but I don't know of such a request.

Takashi

Quoted 7 lines

If not, rejecting major=0 with -EINVAL might be simpler.
If it is kept, the Fixes tag is fine since the patch depends on bd8f21876189, but please mention in the changelog that major=0 never worked before.

Best regards,
Phuc

#3 lore raw

Hi Takashi-san,

Quoted 4 lines
Quoted 4 lines

Note that major=0 has never worked, even before bd8f21876189: the old code treated the positive return value of register_chrdev() as an error and failed with -EIO. bd8f21876189 only changed the symptom.

Quoted 9 lines
Quoted 3 lines

So this patch adds support for a dynamic major rather than fixing a regression. Takashi-san, is a dynamic major something ALSA wants to support?

No, it's supposed to be only the fixed one.

If there is a real demand, we can extend the implementation, but I don't know of such a request.

Thank you for the clarification.

Best regards,
Phuc

#4 lore raw
In reply to Bui Duc Phuc
Quoted 3 lines

Note that major=0 has never worked, even before bd8f21876189: the old code treated the positive return value of register_chrdev() as an error and failed with -EIO. bd8f21876189 only changed the symptom.

Oh, you're right, I didn't notice that. Sorry about that, and thanks.