2011-05-24 10:55:40

by Nicolas Kaiser

Subject: [PATCH] posix clocks: correct error value in posix_clock_poll()

Signed-off-by: Nicolas Kaiser <[email protected]>
---
It looks to me like -ENODEV might not be a good return value
in poll(). Would POLLERR be the correct one in this case?

kernel/time/posix-clock.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/time/posix-clock.c b/kernel/time/posix-clock.c
index c340ca6..2424d3f 100644
--- a/kernel/time/posix-clock.c
+++ b/kernel/time/posix-clock.c
@@ -71,7 +71,7 @@ static unsigned int posix_clock_poll(struct file *fp, poll_table *wait)
int result = 0;

if (!clk)
- return -ENODEV;
+ return POLLERR;

if (clk->ops.poll)
result = clk->ops.poll(clk, fp, wait);
--
1.7.5.rc3

2011-05-24 12:34:36

by Richard Cochran

Subject: Re: [PATCH] posix clocks: correct error value in posix_clock_poll()

On Tue, May 24, 2011 at 12:51:47PM +0200, Nicolas Kaiser wrote:
> Signed-off-by: Nicolas Kaiser <[email protected]>
> ---
> It looks to me like -ENODEV might not be a good return value
> in poll(). Would POLLERR be the correct one in this case?
>
> kernel/time/posix-clock.c | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/kernel/time/posix-clock.c b/kernel/time/posix-clock.c
> index c340ca6..2424d3f 100644
> --- a/kernel/time/posix-clock.c
> +++ b/kernel/time/posix-clock.c
> @@ -71,7 +71,7 @@ static unsigned int posix_clock_poll(struct file *fp, poll_table *wait)
> int result = 0;
>
> if (!clk)
> - return -ENODEV;
> + return POLLERR;

The condition (!clk) is only satisfied when 'zombie' is set,
indicating that the dynamic clock has disappeared. That is why the
file operations uniformly return ENODEV. So, I think it makes sense
the way that it is.

In addition, man 2 poll says,

POLLERR
Error condition (output only).

so using that error code would be misleading, IMHO.

Thanks,
Richard

2011-05-24 14:09:25

by Nicolas Kaiser

Subject: Re: [PATCH] posix clocks: correct error value in posix_clock_poll()

* Richard Cochran <[email protected]>:
> On Tue, May 24, 2011 at 12:51:47PM +0200, Nicolas Kaiser wrote:
> > It looks to me like -ENODEV might not be a good return value
> > in poll(). Would POLLERR be the correct one in this case?

> The condition (!clk) is only satisfied when 'zombie' is set,
> indicating that the dynamic clock has disappeared. That is why the
> file operations uniformly return ENODEV. So, I think it makes sense
> the way that it is.
>
> In addition, man 2 poll says,
>
> POLLERR
> Error condition (output only).
>
> so using that error code would be misleading, IMHO.

I see. Sorry for the noise.

Best regards,
Nicolas Kaiser