The poll condition should only check response_length,
because reads should only be issued if there is data to read.
The response_read flag only prevents double writes.
The problem was that the write set the response_read to false,
enqued a tpm job, and returned. Then application called poll
which checked the response_read flag and returned EPOLLIN.
Then the application called read, but got nothing.
After all that the async_work kicked in.
Added also mutex_lock around the poll check to prevent
other possible race conditions.
Fixes: 9488585b21bef0df12 ("tpm: add support for partial reads")
Reported-by: Mantas Mikulėnas <[email protected]>
Signed-off-by: Tadeusz Struk <[email protected]>
---
drivers/char/tpm/tpm-dev-common.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/char/tpm/tpm-dev-common.c b/drivers/char/tpm/tpm-dev-common.c
index 5eecad233ea1..7312d3214381 100644
--- a/drivers/char/tpm/tpm-dev-common.c
+++ b/drivers/char/tpm/tpm-dev-common.c
@@ -203,12 +203,14 @@ __poll_t tpm_common_poll(struct file *file, poll_table *wait)
__poll_t mask = 0;
poll_wait(file, &priv->async_wait, wait);
+ mutex_lock(&priv->buffer_mutex);
- if (!priv->response_read || priv->response_length)
+ if (priv->response_length)
mask = EPOLLIN | EPOLLRDNORM;
else
mask = EPOLLOUT | EPOLLWRNORM;
+ mutex_unlock(&priv->buffer_mutex);
return mask;
}
On Tue, 2019-03-19 at 13:31 -0700, Tadeusz Struk wrote:
> The poll condition should only check response_length,
> because reads should only be issued if there is data to read.
> The response_read flag only prevents double writes.
> The problem was that the write set the response_read to false,
> enqued a tpm job, and returned. Then application called poll
> which checked the response_read flag and returned EPOLLIN.
> Then the application called read, but got nothing.
> After all that the async_work kicked in.
> Added also mutex_lock around the poll check to prevent
> other possible race conditions.
>
> Fixes: 9488585b21bef0df12 ("tpm: add support for partial reads")
> Reported-by: Mantas Mikulėnas <[email protected]>
> Signed-off-by: Tadeusz Struk <[email protected]>
> ---
> drivers/char/tpm/tpm-dev-common.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/char/tpm/tpm-dev-common.c
> b/drivers/char/tpm/tpm-dev-common.c
> index 5eecad233ea1..7312d3214381 100644
> --- a/drivers/char/tpm/tpm-dev-common.c
> +++ b/drivers/char/tpm/tpm-dev-common.c
> @@ -203,12 +203,14 @@ __poll_t tpm_common_poll(struct file *file,
> poll_table *wait)
> __poll_t mask = 0;
>
> poll_wait(file, &priv->async_wait, wait);
> + mutex_lock(&priv->buffer_mutex);
>
> - if (!priv->response_read || priv->response_length)
> + if (priv->response_length)
> mask = EPOLLIN | EPOLLRDNORM;
> else
> mask = EPOLLOUT | EPOLLWRNORM;
>
> + mutex_unlock(&priv->buffer_mutex);
Just an observation on this: the mutex is now no-longer necessary
because a read on a size_t quantity is always atomic.
James
On 3/20/19 7:30 AM, James Bottomley wrote:
> Just an observation on this: the mutex is now no-longer necessary
> because a read on a size_t quantity is always atomic.
True, that's why it wasn't there at the beginning, but
then things changed and I forgot to add it, so let's
put it there just in case.
Thanks,
--
Tadeusz
On Tue, Mar 19, 2019 at 10:31 PM Tadeusz Struk <[email protected]> wrote:
>
> The poll condition should only check response_length,
> because reads should only be issued if there is data to read.
> The response_read flag only prevents double writes.
> The problem was that the write set the response_read to false,
> enqued a tpm job, and returned. Then application called poll
> which checked the response_read flag and returned EPOLLIN.
> Then the application called read, but got nothing.
> After all that the async_work kicked in.
> Added also mutex_lock around the poll check to prevent
> other possible race conditions.
>
> Fixes: 9488585b21bef0df12 ("tpm: add support for partial reads")
> Reported-by: Mantas Mikulėnas <[email protected]>
> Signed-off-by: Tadeusz Struk <[email protected]>
> ---
> drivers/char/tpm/tpm-dev-common.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/char/tpm/tpm-dev-common.c b/drivers/char/tpm/tpm-dev-common.c
> index 5eecad233ea1..7312d3214381 100644
> --- a/drivers/char/tpm/tpm-dev-common.c
> +++ b/drivers/char/tpm/tpm-dev-common.c
> @@ -203,12 +203,14 @@ __poll_t tpm_common_poll(struct file *file, poll_table *wait)
> __poll_t mask = 0;
>
> poll_wait(file, &priv->async_wait, wait);
> + mutex_lock(&priv->buffer_mutex);
>
> - if (!priv->response_read || priv->response_length)
> + if (priv->response_length)
> mask = EPOLLIN | EPOLLRDNORM;
> else
> mask = EPOLLOUT | EPOLLWRNORM;
>
> + mutex_unlock(&priv->buffer_mutex);
> return mask;
> }
Thanks, this patch seems to work, and I apologize for not responding
to test the patches earlier.
Any chance it'll be submitted for stable 5.0.x as well?
--
Mantas Mikulėnas
On 3/20/19 11:51 AM, Mantas Mikulėnas wrote:
> Thanks, this patch seems to work, and I apologize for not responding
> to test the patches earlier.
Thanks for testing.
>
> Any chance it'll be submitted for stable 5.0.x as well?
Yes, it's a regression. I included the "Fixes" tag so
it should be applied to all affected versions.
In this case it's 5.0 only.
Thanks,
--
Tadeusz
On Tue, Mar 19, 2019 at 01:31:34PM -0700, Tadeusz Struk wrote:
> The poll condition should only check response_length,
> because reads should only be issued if there is data to read.
> The response_read flag only prevents double writes.
> The problem was that the write set the response_read to false,
> enqued a tpm job, and returned. Then application called poll
> which checked the response_read flag and returned EPOLLIN.
> Then the application called read, but got nothing.
> After all that the async_work kicked in.
> Added also mutex_lock around the poll check to prevent
> other possible race conditions.
>
> Fixes: 9488585b21bef0df12 ("tpm: add support for partial reads")
> Reported-by: Mantas Mikulėnas <[email protected]>
> Signed-off-by: Tadeusz Struk <[email protected]>
Can you send v3 with the comment and cc to stable? Then this
should be fine.
/Jarkko
On Wed, Mar 20, 2019 at 12:18:42PM -0700, Tadeusz Struk wrote:
> On 3/20/19 11:51 AM, Mantas Mikulėnas wrote:
> > Thanks, this patch seems to work, and I apologize for not responding
> > to test the patches earlier.
>
> Thanks for testing.
>
> >
> > Any chance it'll be submitted for stable 5.0.x as well?
>
> Yes, it's a regression. I included the "Fixes" tag so
> it should be applied to all affected versions.
> In this case it's 5.0 only.
> Thanks,
> --
> Tadeusz
Add a tested-by for the next version.
/Jarkko