| Message ID | 20200320104031.31701-1-alexandru.ardelean@analog.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [RFC] iio: buffer: Don't allow buffers without any channels enabled to be activated | expand |
On 3/20/20 11:40 AM, Alexandru Ardelean wrote: > From: Lars-Peter Clausen <lars@metafoo.de> > > Before activating a buffer make sure that at least one channel is enabled. > Activating a buffer with 0 channels enabled doesn't make too much sense and > disallowing this case makes sure that individual driver don't have to add > special case code to handle it. > > Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> > Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com> > --- > > * Found this relic-patch in our tree, from 6 years ago: > https://github.com/analogdevicesinc/linux/commit/6d680e49d459c > It got moved around a bit, and this is the current form in the ADI tree. > So, this is also a bit of an RFC, but if the idea is valid, maybe it's > worth considering upstream. I don't know of any arguments against it, > but I could be surprised. Hm, a bit weird that this one never made it upstream considering how simple it is. Did you check that the issue still occurs? I can't see anything in the code that prevents it, but who knows, maybe it was fixed by something else. > > drivers/iio/industrialio-buffer.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/drivers/iio/industrialio-buffer.c b/drivers/iio/industrialio-buffer.c > index 4ada5592aa2b..f222a118d0d3 100644 > --- a/drivers/iio/industrialio-buffer.c > +++ b/drivers/iio/industrialio-buffer.c > @@ -1031,6 +1031,12 @@ static int __iio_update_buffers(struct iio_dev *indio_dev, > return ret; > > if (insert_buffer) { > + if (bitmap_empty(insert_buffer->scan_mask, > + indio_dev->masklength)) { > + ret = -EINVAL; > + goto err_free_config; > + } Since the check is so simple it might make sense to do it as the very first thing before iio_verify_update(). > + > ret = iio_buffer_request_update(indio_dev, insert_buffer); > if (ret) > goto err_free_config; >
On Fri, 2020-03-20 at 11:55 +0100, Lars-Peter Clausen wrote: > On 3/20/20 11:40 AM, Alexandru Ardelean wrote: > > From: Lars-Peter Clausen <lars@metafoo.de> > > > > Before activating a buffer make sure that at least one channel is enabled. > > Activating a buffer with 0 channels enabled doesn't make too much sense and > > disallowing this case makes sure that individual driver don't have to add > > special case code to handle it. > > > > Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> > > Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com> > > --- > > > > * Found this relic-patch in our tree, from 6 years ago: > > https://github.com/analogdevicesinc/linux/commit/6d680e49d459c > > It got moved around a bit, and this is the current form in the ADI tree. > > So, this is also a bit of an RFC, but if the idea is valid, maybe it's > > worth considering upstream. I don't know of any arguments against it, > > but I could be surprised. > > Hm, a bit weird that this one never made it upstream considering how > simple it is. > > Did you check that the issue still occurs? I can't see anything in the > code that prevents it, but who knows, maybe it was fixed by something else. i did not think to check behavior/issues; i'll try to make some time for that; i caught this one while diff-ing the upstream & ADI trees, and i needed to dig a bit more into the ADI git history on it; i was a bit puzzled for a while, because some rework patches were upstreamed without this patch: https://lore.kernel.org/linux-iio/55585CAA.6000506@kernel.org/ https://lore.kernel.org/linux-iio/5560685A.5060504@kernel.org/ i also did not find any discussions/upstream attempt for this patch particularly so, it was easier for me just to RFC this > > > drivers/iio/industrialio-buffer.c | 6 ++++++ > > 1 file changed, 6 insertions(+) > > > > diff --git a/drivers/iio/industrialio-buffer.c b/drivers/iio/industrialio- > > buffer.c > > index 4ada5592aa2b..f222a118d0d3 100644 > > --- a/drivers/iio/industrialio-buffer.c > > +++ b/drivers/iio/industrialio-buffer.c > > @@ -1031,6 +1031,12 @@ static int __iio_update_buffers(struct iio_dev > > *indio_dev, > > return ret; > > > > if (insert_buffer) { > > + if (bitmap_empty(insert_buffer->scan_mask, > > + indio_dev->masklength)) { > > + ret = -EINVAL; > > + goto err_free_config; > > + } > > Since the check is so simple it might make sense to do it as the very > first thing before iio_verify_update(). works for me; > > > + > > ret = iio_buffer_request_update(indio_dev, insert_buffer); > > if (ret) > > goto err_free_config; > >
On Fri, 20 Mar 2020 11:16:12 +0000 "Ardelean, Alexandru" <alexandru.Ardelean@analog.com> wrote: > On Fri, 2020-03-20 at 11:55 +0100, Lars-Peter Clausen wrote: > > On 3/20/20 11:40 AM, Alexandru Ardelean wrote: > > > From: Lars-Peter Clausen <lars@metafoo.de> > > > > > > Before activating a buffer make sure that at least one channel is enabled. > > > Activating a buffer with 0 channels enabled doesn't make too much sense and > > > disallowing this case makes sure that individual driver don't have to add > > > special case code to handle it. > > > > > > Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> > > > Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com> > > > --- > > > > > > * Found this relic-patch in our tree, from 6 years ago: > > > https://github.com/analogdevicesinc/linux/commit/6d680e49d459c > > > It got moved around a bit, and this is the current form in the ADI tree. > > > So, this is also a bit of an RFC, but if the idea is valid, maybe it's > > > worth considering upstream. I don't know of any arguments against it, > > > but I could be surprised. > > > > Hm, a bit weird that this one never made it upstream considering how > > simple it is. > > > > Did you check that the issue still occurs? I can't see anything in the > > code that prevents it, but who knows, maybe it was fixed by something else. > > i did not think to check behavior/issues; > i'll try to make some time for that; I can't immediately think of anything that would stop this case. However, good if you could confirm it. (I don't have a setup running right now to test against) > i caught this one while diff-ing the upstream & ADI trees, and i needed to dig a > bit more into the ADI git history on it; > > i was a bit puzzled for a while, because some rework patches were upstreamed > without this patch: > > https://lore.kernel.org/linux-iio/55585CAA.6000506@kernel.org/ > https://lore.kernel.org/linux-iio/5560685A.5060504@kernel.org/ > > i also did not find any discussions/upstream attempt for this patch particularly > > so, it was easier for me just to RFC this > > > > > > drivers/iio/industrialio-buffer.c | 6 ++++++ > > > 1 file changed, 6 insertions(+) > > > > > > diff --git a/drivers/iio/industrialio-buffer.c b/drivers/iio/industrialio- > > > buffer.c > > > index 4ada5592aa2b..f222a118d0d3 100644 > > > --- a/drivers/iio/industrialio-buffer.c > > > +++ b/drivers/iio/industrialio-buffer.c > > > @@ -1031,6 +1031,12 @@ static int __iio_update_buffers(struct iio_dev > > > *indio_dev, > > > return ret; > > > > > > if (insert_buffer) { > > > + if (bitmap_empty(insert_buffer->scan_mask, > > > + indio_dev->masklength)) { > > > + ret = -EINVAL; > > > + goto err_free_config; > > > + } > > > > Since the check is so simple it might make sense to do it as the very > > first thing before iio_verify_update(). > > works for me; > > > > > > + > > > ret = iio_buffer_request_update(indio_dev, insert_buffer); > > > if (ret) > > > goto err_free_config; > > >
On Sat, Mar 21, 2020 at 8:22 PM Jonathan Cameron <jic23@kernel.org> wrote: > On Fri, 20 Mar 2020 11:16:12 +0000 > "Ardelean, Alexandru" <alexandru.Ardelean@analog.com> wrote: > > On Fri, 2020-03-20 at 11:55 +0100, Lars-Peter Clausen wrote: > > > On 3/20/20 11:40 AM, Alexandru Ardelean wrote: > > > > + if (bitmap_empty(insert_buffer->scan_mask, > > > > + indio_dev->masklength)) { I guess it's much better to put on one line (despite possible over 80 limit).
On Sat, 2020-03-21 at 18:21 +0000, Jonathan Cameron wrote: > On Fri, 20 Mar 2020 11:16:12 +0000 > "Ardelean, Alexandru" <alexandru.Ardelean@analog.com> wrote: > > > On Fri, 2020-03-20 at 11:55 +0100, Lars-Peter Clausen wrote: > > > On 3/20/20 11:40 AM, Alexandru Ardelean wrote: > > > > From: Lars-Peter Clausen <lars@metafoo.de> > > > > > > > > Before activating a buffer make sure that at least one channel is > > > > enabled. > > > > Activating a buffer with 0 channels enabled doesn't make too much sense > > > > and > > > > disallowing this case makes sure that individual driver don't have to > > > > add > > > > special case code to handle it. > > > > > > > > Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> > > > > Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com> > > > > --- > > > > > > > > * Found this relic-patch in our tree, from 6 years ago: > > > > https://github.com/analogdevicesinc/linux/commit/6d680e49d459c > > > > It got moved around a bit, and this is the current form in the ADI > > > > tree. > > > > So, this is also a bit of an RFC, but if the idea is valid, maybe > > > > it's > > > > worth considering upstream. I don't know of any arguments against it, > > > > but I could be surprised. > > > > > > Hm, a bit weird that this one never made it upstream considering how > > > simple it is. > > > > > > Did you check that the issue still occurs? I can't see anything in the > > > code that prevents it, but who knows, maybe it was fixed by something > > > else. > > > > i did not think to check behavior/issues; > > i'll try to make some time for that; > > I can't immediately think of anything that would stop this case. > > However, good if you could confirm it. (I don't have a setup running > right now to test against) I've instrumented the code a bit. So, with this change [moved in iio_verify_update() as Lars suggested]: root@analog:~# cd /sys/bus/iio/devices/iio\:device3/buffer root@analog:/sys/bus/iio/devices/iio:device3/buffer# echo 1 > enable 000000000 iio_verify_update 748 indio_dev->masklength 2 *insert_buffer->scan_mask 00000000 11111111 iio_verify_update 753 222222 __iio_update_buffers 1115 ret -22 333333333 iio_buffer_store_enable 1241 ret -22 -bash: echo: write error: Invalid argument so, 11111111, 222222 & 333333333 are all error paths. Without the patch: root@analog:~# cd /sys/bus/iio/devices/iio\:device3/buffer root@analog:/sys/bus/iio/devices/iio:device3/buffer# echo 1 > enable 000000000 iio_verify_update 748 indio_dev->masklength 2 *insert_buffer->scan_mask 00000000 root@analog:/sys/bus/iio/devices/iio:device3/buffer# no error path is hit; error paths are still there, but the bitmap_empty() check removed; So, buffer gets enabled, but scan_mask is empty. Will follow-up a V2 on this and attach this information. > > > > i caught this one while diff-ing the upstream & ADI trees, and i needed to > > dig a > > bit more into the ADI git history on it; > > > > i was a bit puzzled for a while, because some rework patches were upstreamed > > without this patch: > > > > https://lore.kernel.org/linux-iio/55585CAA.6000506@kernel.org/ > > https://lore.kernel.org/linux-iio/5560685A.5060504@kernel.org/ > > > > i also did not find any discussions/upstream attempt for this patch > > particularly > > > > so, it was easier for me just to RFC this > > > > > > > > > drivers/iio/industrialio-buffer.c | 6 ++++++ > > > > 1 file changed, 6 insertions(+) > > > > > > > > diff --git a/drivers/iio/industrialio-buffer.c > > > > b/drivers/iio/industrialio- > > > > buffer.c > > > > index 4ada5592aa2b..f222a118d0d3 100644 > > > > --- a/drivers/iio/industrialio-buffer.c > > > > +++ b/drivers/iio/industrialio-buffer.c > > > > @@ -1031,6 +1031,12 @@ static int __iio_update_buffers(struct iio_dev > > > > *indio_dev, > > > > return ret; > > > > > > > > if (insert_buffer) { > > > > + if (bitmap_empty(insert_buffer->scan_mask, > > > > + indio_dev->masklength)) { > > > > + ret = -EINVAL; > > > > + goto err_free_config; > > > > + } > > > > > > Since the check is so simple it might make sense to do it as the very > > > first thing before iio_verify_update(). > > > > works for me; > > > > > > > > > + > > > > ret = iio_buffer_request_update(indio_dev, > > > > insert_buffer); > > > > if (ret) > > > > goto err_free_config; > > > >
On Wed, 2020-03-25 at 08:17 +0000, Ardelean, Alexandru wrote: > On Sat, 2020-03-21 at 18:21 +0000, Jonathan Cameron wrote: > > On Fri, 20 Mar 2020 11:16:12 +0000 > > "Ardelean, Alexandru" <alexandru.Ardelean@analog.com> wrote: > > > > > On Fri, 2020-03-20 at 11:55 +0100, Lars-Peter Clausen wrote: > > > > On 3/20/20 11:40 AM, Alexandru Ardelean wrote: > > > > > From: Lars-Peter Clausen <lars@metafoo.de> > > > > > > > > > > Before activating a buffer make sure that at least one channel is > > > > > enabled. > > > > > Activating a buffer with 0 channels enabled doesn't make too much > > > > > sense > > > > > and > > > > > disallowing this case makes sure that individual driver don't have to > > > > > add > > > > > special case code to handle it. > > > > > > > > > > Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> > > > > > Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com> > > > > > --- > > > > > > > > > > * Found this relic-patch in our tree, from 6 years ago: > > > > > https://github.com/analogdevicesinc/linux/commit/6d680e49d459c > > > > > It got moved around a bit, and this is the current form in the ADI > > > > > tree. > > > > > So, this is also a bit of an RFC, but if the idea is valid, maybe > > > > > it's > > > > > worth considering upstream. I don't know of any arguments against > > > > > it, > > > > > but I could be surprised. > > > > > > > > Hm, a bit weird that this one never made it upstream considering how > > > > simple it is. > > > > > > > > Did you check that the issue still occurs? I can't see anything in the > > > > code that prevents it, but who knows, maybe it was fixed by something > > > > else. > > > > > > i did not think to check behavior/issues; > > > i'll try to make some time for that; > > > > I can't immediately think of anything that would stop this case. > > > > However, good if you could confirm it. (I don't have a setup running > > right now to test against) > > I've instrumented the code a bit. > So, with this change [moved in iio_verify_update() as Lars suggested]: > > root@analog:~# cd /sys/bus/iio/devices/iio\:device3/buffer > root@analog:/sys/bus/iio/devices/iio:device3/buffer# echo 1 > enable > 000000000 iio_verify_update 748 > indio_dev->masklength 2 > *insert_buffer->scan_mask 00000000 > 11111111 iio_verify_update 753 > 222222 __iio_update_buffers 1115 ret -22 > 333333333 iio_buffer_store_enable 1241 ret -22 > -bash: echo: write error: Invalid argument > > so, 11111111, 222222 & 333333333 are all error paths. > > Without the patch: > > root@analog:~# cd /sys/bus/iio/devices/iio\:device3/buffer > root@analog:/sys/bus/iio/devices/iio:device3/buffer# echo 1 > enable > 000000000 iio_verify_update 748 > indio_dev->masklength 2 > *insert_buffer->scan_mask 00000000 > root@analog:/sys/bus/iio/devices/iio:device3/buffer# > > no error path is hit; > error paths are still there, but the bitmap_empty() check removed; > > So, buffer gets enabled, but scan_mask is empty. > > Will follow-up a V2 on this and attach this information. Right. As a quick follow-up here: if the scan_mask is not empty [i.e. at least one scan_element is enabled] all works fine. > > > > > > > i caught this one while diff-ing the upstream & ADI trees, and i needed to > > > dig a > > > bit more into the ADI git history on it; > > > > > > i was a bit puzzled for a while, because some rework patches were > > > upstreamed > > > without this patch: > > > > > > https://lore.kernel.org/linux-iio/55585CAA.6000506@kernel.org/ > > > https://lore.kernel.org/linux-iio/5560685A.5060504@kernel.org/ > > > > > > i also did not find any discussions/upstream attempt for this patch > > > particularly > > > > > > so, it was easier for me just to RFC this > > > > > > > > > > > > drivers/iio/industrialio-buffer.c | 6 ++++++ > > > > > 1 file changed, 6 insertions(+) > > > > > > > > > > diff --git a/drivers/iio/industrialio-buffer.c > > > > > b/drivers/iio/industrialio- > > > > > buffer.c > > > > > index 4ada5592aa2b..f222a118d0d3 100644 > > > > > --- a/drivers/iio/industrialio-buffer.c > > > > > +++ b/drivers/iio/industrialio-buffer.c > > > > > @@ -1031,6 +1031,12 @@ static int __iio_update_buffers(struct iio_dev > > > > > *indio_dev, > > > > > return ret; > > > > > > > > > > if (insert_buffer) { > > > > > + if (bitmap_empty(insert_buffer->scan_mask, > > > > > + indio_dev->masklength)) { > > > > > + ret = -EINVAL; > > > > > + goto err_free_config; > > > > > + } > > > > > > > > Since the check is so simple it might make sense to do it as the very > > > > first thing before iio_verify_update(). > > > > > > works for me; > > > > > > > > > > > > + > > > > > ret = iio_buffer_request_update(indio_dev, > > > > > insert_buffer); > > > > > if (ret) > > > > > goto err_free_config; > > > > >
diff --git a/drivers/iio/industrialio-buffer.c b/drivers/iio/industrialio-buffer.c index 4ada5592aa2b..f222a118d0d3 100644 --- a/drivers/iio/industrialio-buffer.c +++ b/drivers/iio/industrialio-buffer.c @@ -1031,6 +1031,12 @@ static int __iio_update_buffers(struct iio_dev *indio_dev, return ret; if (insert_buffer) { + if (bitmap_empty(insert_buffer->scan_mask, + indio_dev->masklength)) { + ret = -EINVAL; + goto err_free_config; + } + ret = iio_buffer_request_update(indio_dev, insert_buffer); if (ret) goto err_free_config;