Re: [Y2038] [PATCH v7 6/9] ALSA: Avoid using timespec for struct snd_timer_tread

On Thu, 2019-12-12 at 10:57 +0100, Arnd Bergmann wrote:

On Thu, Dec 12, 2019 at 1:14 AM Ben Hutchings ben.hutchings@codethink.co.uk wrote:

...

On Wed, 2019-12-11 at 22:20 +0100, Arnd Bergmann wrote: [...]

...

+static int snd_timer_user_tread(void __user *argp, struct snd_timer_user *tu,

•                         unsigned int cmd, bool compat)
  

+{

• int __user *p = argp;
  

• int xarg, old_tread;
  

• if (tu->timeri) /* too late */
  

•         return -EBUSY;
  

• if (get_user(xarg, p))
  

•         return -EFAULT;
  

• old_tread = tu->tread;
  

• if (!xarg)
  

•         tu->tread = TREAD_FORMAT_NONE;
  

• else if (cmd == SNDRV_TIMER_IOCTL_TREAD64 ||
  

•          (IS_ENABLED(CONFIG_64BITS) && !compat))
  

This needs to check for CONFIG_64BIT not CONFIG_64BITS.

Fixed now, good catch!

...

...

@@ -2145,14 +2202,34 @@ static ssize_t snd_timer_user_read(struct file *file, char __user *buffer,

•         case TREAD_FORMAT_NONE:
                  if (copy_to_user(buffer, &tu->queue[qhead],
                                   sizeof(struct snd_timer_read)))
                          err = -EFAULT;
  

•                 break;
  

•         default:
  

•                 err = -ENOTSUPP;
  

[...]

This is not a valid error code for returning to user-space, but this case should be impossible so I don't think it matters.

Agreed. Maybe it should also WARN_ON(1), as there getting here would indicate a bug in the kernel.

Yes, WARN_ON() or WARN_ON_ONCE() would make sense.

Ben.