| Message ID | 20200722155103.979802-9-jic23@kernel.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | IIO: Fused set 1 and 2 of timestamp alignment fixes | expand |
On Wed, Jul 22, 2020 at 6:53 PM Jonathan Cameron <jic23@kernel.org> wrote: > > From: Jonathan Cameron <Jonathan.Cameron@huawei.com> > > One of a class of bugs pointed out by Lars in a recent review. > iio_push_to_buffers_with_timestamp assumes the buffer used is aligned > to the size of the timestamp (8 bytes). This is not guaranteed in > this driver which uses a 24 byte array of smaller elements on the stack. > As Lars also noted this anti pattern can involve a leak of data to > userspace and that indeed can happen here. We close both issues by > moving to a suitable array in the iio_priv() data with alignment > explicitly requested. This data is allocated with kzalloc so no > data can leak appart from previous readings. > > Depending on the enabled channels, the location of the timestamp > can be at various aligned offsets through the buffer. As such we > any use of a structure to enforce this alignment would incorrectly > suggest a single location for the timestamp. ... > + /* Ensure timestamp will be naturally aligned if present */ > + u8 buffer[24] __aligned(8); Why can't we use proper structure here? > @@ -445,7 +447,6 @@ static irqreturn_t si1145_trigger_handler(int irq, void *private) > * 6*2 bytes channels data + 4 bytes alignment + > * 8 bytes timestamp > */ > - u8 buffer[24]; Seems even the old comment shows how it should look like...
On Wed, Jul 22, 2020 at 10:43 PM Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > > On Wed, Jul 22, 2020 at 6:53 PM Jonathan Cameron <jic23@kernel.org> wrote: > > > > From: Jonathan Cameron <Jonathan.Cameron@huawei.com> > > > > One of a class of bugs pointed out by Lars in a recent review. > > iio_push_to_buffers_with_timestamp assumes the buffer used is aligned > > to the size of the timestamp (8 bytes). This is not guaranteed in > > this driver which uses a 24 byte array of smaller elements on the stack. > > As Lars also noted this anti pattern can involve a leak of data to > > userspace and that indeed can happen here. We close both issues by > > moving to a suitable array in the iio_priv() data with alignment > > explicitly requested. This data is allocated with kzalloc so no > > data can leak appart from previous readings. > > > > Depending on the enabled channels, the location of the timestamp > > can be at various aligned offsets through the buffer. As such we > > any use of a structure to enforce this alignment would incorrectly > > suggest a single location for the timestamp. > > ... > > > + /* Ensure timestamp will be naturally aligned if present */ > > + u8 buffer[24] __aligned(8); > > Why can't we use proper structure here? > > > @@ -445,7 +447,6 @@ static irqreturn_t si1145_trigger_handler(int irq, void *private) > > * 6*2 bytes channels data + 4 bytes alignment + > > * 8 bytes timestamp > > */ > > - u8 buffer[24]; > > Seems even the old comment shows how it should look like... I think I understand now. Basically it's a dynamic amount of channels (up to 6) before you get a timestamp.
On Wed, 22 Jul 2020 22:45:59 +0300 Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > On Wed, Jul 22, 2020 at 10:43 PM Andy Shevchenko > <andy.shevchenko@gmail.com> wrote: > > > > On Wed, Jul 22, 2020 at 6:53 PM Jonathan Cameron <jic23@kernel.org> wrote: > > > > > > From: Jonathan Cameron <Jonathan.Cameron@huawei.com> > > > > > > One of a class of bugs pointed out by Lars in a recent review. > > > iio_push_to_buffers_with_timestamp assumes the buffer used is aligned > > > to the size of the timestamp (8 bytes). This is not guaranteed in > > > this driver which uses a 24 byte array of smaller elements on the stack. > > > As Lars also noted this anti pattern can involve a leak of data to > > > userspace and that indeed can happen here. We close both issues by > > > moving to a suitable array in the iio_priv() data with alignment > > > explicitly requested. This data is allocated with kzalloc so no > > > data can leak appart from previous readings. > > > > > > Depending on the enabled channels, the location of the timestamp > > > can be at various aligned offsets through the buffer. As such we > > > any use of a structure to enforce this alignment would incorrectly > > > suggest a single location for the timestamp. > > > > ... > > > > > + /* Ensure timestamp will be naturally aligned if present */ > > > + u8 buffer[24] __aligned(8); > > > > Why can't we use proper structure here? > > > > > @@ -445,7 +447,6 @@ static irqreturn_t si1145_trigger_handler(int irq, void *private) > > > * 6*2 bytes channels data + 4 bytes alignment + > > > * 8 bytes timestamp > > > */ > > > - u8 buffer[24]; > > > > Seems even the old comment shows how it should look like... > > I think I understand now. Basically it's a dynamic amount of channels > (up to 6) before you get a timestamp. > Exactly. Comment is giving the largest it can be, not what is needed for a given configuration of the device. Should indeed drop that comment. Obviously went into automated mode and stopped actually reading what was in front of me. Jonathan
On Thu, 23 Jul 2020 12:25:17 +0100 Jonathan Cameron <Jonathan.Cameron@Huawei.com> wrote: > On Wed, 22 Jul 2020 22:45:59 +0300 > Andy Shevchenko <andy.shevchenko@gmail.com> wrote: > > > On Wed, Jul 22, 2020 at 10:43 PM Andy Shevchenko > > <andy.shevchenko@gmail.com> wrote: > > > > > > On Wed, Jul 22, 2020 at 6:53 PM Jonathan Cameron <jic23@kernel.org> wrote: > > > > > > > > From: Jonathan Cameron <Jonathan.Cameron@huawei.com> > > > > > > > > One of a class of bugs pointed out by Lars in a recent review. > > > > iio_push_to_buffers_with_timestamp assumes the buffer used is aligned > > > > to the size of the timestamp (8 bytes). This is not guaranteed in > > > > this driver which uses a 24 byte array of smaller elements on the stack. > > > > As Lars also noted this anti pattern can involve a leak of data to > > > > userspace and that indeed can happen here. We close both issues by > > > > moving to a suitable array in the iio_priv() data with alignment > > > > explicitly requested. This data is allocated with kzalloc so no > > > > data can leak appart from previous readings. > > > > > > > > Depending on the enabled channels, the location of the timestamp > > > > can be at various aligned offsets through the buffer. As such we > > > > any use of a structure to enforce this alignment would incorrectly > > > > suggest a single location for the timestamp. > > > > > > ... > > > > > > > + /* Ensure timestamp will be naturally aligned if present */ > > > > + u8 buffer[24] __aligned(8); > > > > > > Why can't we use proper structure here? > > > > > > > @@ -445,7 +447,6 @@ static irqreturn_t si1145_trigger_handler(int irq, void *private) > > > > * 6*2 bytes channels data + 4 bytes alignment + > > > > * 8 bytes timestamp > > > > */ > > > > - u8 buffer[24]; > > > > > > Seems even the old comment shows how it should look like... > > > > I think I understand now. Basically it's a dynamic amount of channels > > (up to 6) before you get a timestamp. > > > Exactly. Comment is giving the largest it can be, not what is needed for > a given configuration of the device. > > Should indeed drop that comment. Obviously went into automated mode and stopped > actually reading what was in front of me. I've adjusted the comment as requested by Andy (and moved it!). Fits under Andy's class 2 so applied to the togreg branch of iio.git and marked for stable. No great rush for this, beyond the fact that I'll keep forgetting to actually sort these out! Thanks, Jonathan > > Jonathan > >
diff --git a/drivers/iio/light/si1145.c b/drivers/iio/light/si1145.c index 155faaea8c72..e8bdc221d65b 100644 --- a/drivers/iio/light/si1145.c +++ b/drivers/iio/light/si1145.c @@ -179,6 +179,8 @@ struct si1145_data { bool autonomous; struct iio_trigger *trig; int meas_rate; + /* Ensure timestamp will be naturally aligned if present */ + u8 buffer[24] __aligned(8); }; /* @@ -445,7 +447,6 @@ static irqreturn_t si1145_trigger_handler(int irq, void *private) * 6*2 bytes channels data + 4 bytes alignment + * 8 bytes timestamp */ - u8 buffer[24]; int i, j = 0; int ret; u8 irq_status = 0; @@ -478,7 +479,7 @@ static irqreturn_t si1145_trigger_handler(int irq, void *private) ret = i2c_smbus_read_i2c_block_data_or_emulated( data->client, indio_dev->channels[i].address, - sizeof(u16) * run, &buffer[j]); + sizeof(u16) * run, &data->buffer[j]); if (ret < 0) goto done; j += run * sizeof(u16); @@ -493,7 +494,7 @@ static irqreturn_t si1145_trigger_handler(int irq, void *private) goto done; } - iio_push_to_buffers_with_timestamp(indio_dev, buffer, + iio_push_to_buffers_with_timestamp(indio_dev, data->buffer, iio_get_time_ns(indio_dev)); done: