Skip to contentSkip to conversation

different return value from some ioctl commands for x86 binary executed on x86_64

DISCUSSION Takashi Iwai 2 participants 2 messages Mar 17, 2016 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw
Quoted 104 lines · Takashi Sakamoto

Hi,

When testing ioctl compatibility layer for ALSA ctl interface, I found that error code of some ioctl commands are different depending on binary architecture.

The commands are SNDRV_CTL_IOCTL_ELEM_READ/SNDRV_CTL_IOCTL_ELEM_WRITE. When passing data for non-existent element, ENOENT is returned by the core implementation. On the other hand, ENXIO is returned by the compatibility layer.

This code is for reproduction of this bug.

#include <stdio.h>
#include <stdlib.h>
#include <stdbool.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <errno.h>
#include <string.h>

#include <sound/asound.h>

static void check_elem_read(int fd)

{
        struct snd_ctl_elem_value data = {0};
        if (ioctl(fd, SNDRV_CTL_IOCTL_ELEM_READ, &data) < 0) {
                printf("ioctl(SNDRV_CTL_IOCTL_ELEM_READ): %s\n",
                       strerror(errno));
        }
}

static void check_elem_write(int fd)

{
        struct snd_ctl_elem_value data = {0};
        if (ioctl(fd, SNDRV_CTL_IOCTL_ELEM_WRITE, &data) < 0) {
                printf("ioctl(SNDRV_CTL_IOCTL_ELEM_WRITE): %s\n",
                       strerror(errno));
        }
}

int main(void)

{
        int fd;

        fd = open("/dev/snd/controlC0", O_RDONLY);
        if (fd < 0) {
                printf("open(2): %s\n", strerror(errno));
                return EXIT_FAILURE;
        }

        check_elem_read(fd);
        check_elem_write(fd);

        return EXIT_SUCCESS;
}

I tested on x86_64 machine with enough support for multi-arch.

$ gcc -Wall -m64 -o ./test ./test.c ; ./test
ioctl(SNDRV_CTL_IOCTL_ELEM_READ): No such file or directory
ioctl(SNDRV_CTL_IOCTL_ELEM_WRITE): No such file or directory
$ gcc -Wall -m32 -o ./test ./test.c ; ./test
ioctl(SNDRV_CTL_IOCTL_ELEM_READ): No such device or address
ioctl(SNDRV_CTL_IOCTL_ELEM_WRITE): No such device or address

This bug is due to return value of condition statement for 'snd_ctl_find()'. I wrote a patch to fix it.

Quoted 1 line

From d7a8fd14a1bc100dbaf9bf28af47094d5ccb882b Mon Sep 17 00:00:00 2001

From: Takashi Sakamoto <[email protected]>
Date: Wed, 16 Mar 2016 13:40:31 +0900
Subject: [PATCH] ALSA: ctl: change error code in compatibility layer so that
it's the same code as core implementation

In control compatibility layer, when no elements are found by ELEM_READ/ELEM_WRITE ioctl commands, ENXIO is returned. On the other hand, in core implementation, ENOENT is returned. This is not good for applications.

This commit changes the return value from the compatibility layer so that the same error code is returned.

sound/core/control_compat.c +1 -1
index 0608f21..1fa7076 100644
@@ -196,7 +196,7 @@ static int get_ctl_type(struct snd_card *card,
 struct snd_ctl_elem_id *id,
        kctl = snd_ctl_find_id(card, id);
        if (! kctl) {
                up_read(&card->controls_rwsem);
-               return -ENXIO;
+               return -ENOENT;
        }
        info = kzalloc(sizeof(*info), GFP_KERNEL);
        if (info == NULL) {
Quoted 3 lines

Well, it's already in merge window for Linux 4.6. I don't mind to postpone this patch to next developing cycle, while I also think it better to fix the bug now. So I leave the decision to maintainers.

Looks good to me, and trivial enough to be merged for 4.6. Please resubmit with a proper sign off.

thanks,

Takashi

#2 lore raw
Quoted 62 lines · Takashi Iwai
Quoted 60 lines · Takashi Sakamoto

Hi,

When testing ioctl compatibility layer for ALSA ctl interface, I found that error code of some ioctl commands are different depending on binary architecture.

The commands are SNDRV_CTL_IOCTL_ELEM_READ/SNDRV_CTL_IOCTL_ELEM_WRITE. When passing data for non-existent element, ENOENT is returned by the core implementation. On the other hand, ENXIO is returned by the compatibility layer.

This code is for reproduction of this bug.

(snip)

I tested on x86_64 machine with enough support for multi-arch.

$ gcc -Wall -m64 -o ./test ./test.c ; ./test
ioctl(SNDRV_CTL_IOCTL_ELEM_READ): No such file or directory
ioctl(SNDRV_CTL_IOCTL_ELEM_WRITE): No such file or directory
$ gcc -Wall -m32 -o ./test ./test.c ; ./test
ioctl(SNDRV_CTL_IOCTL_ELEM_READ): No such device or address
ioctl(SNDRV_CTL_IOCTL_ELEM_WRITE): No such device or address

This bug is due to return value of condition statement for 'snd_ctl_find()'. I wrote a patch to fix it.

Quoted 1 line

From d7a8fd14a1bc100dbaf9bf28af47094d5ccb882b Mon Sep 17 00:00:00 2001

From: Takashi Sakamoto <[email protected]>
Date: Wed, 16 Mar 2016 13:40:31 +0900
Subject: [PATCH] ALSA: ctl: change error code in compatibility layer so that
it's the same code as core implementation

In control compatibility layer, when no elements are found by ELEM_READ/ELEM_WRITE ioctl commands, ENXIO is returned. On the other hand, in core implementation, ENOENT is returned. This is not good for applications.

This commit changes the return value from the compatibility layer so that the same error code is returned.

sound/core/control_compat.c +1 -1
index 0608f21..1fa7076 100644
@@ -196,7 +196,7 @@ static int get_ctl_type(struct snd_card *card,
 struct snd_ctl_elem_id *id,
        kctl = snd_ctl_find_id(card, id);
        if (! kctl) {
                up_read(&card->controls_rwsem);
-               return -ENXIO;
+               return -ENOENT;
        }
        info = kzalloc(sizeof(*info), GFP_KERNEL);
        if (info == NULL) {
Quoted 6 lines
Quoted 3 lines

Well, it's already in merge window for Linux 4.6. I don't mind to postpone this patch to next developing cycle, while I also think it better to fix the bug now. So I leave the decision to maintainers.

Looks good to me, and trivial enough to be merged for 4.6. Please resubmit with a proper sign off.

OK, thanks.

Takashi Sakamoto