| Message ID | 20170921113532.15297-1-peter.ujfalusi@ti.com |
|---|---|
| State | New |
| Headers | show |
| Series | dmaengine: ti-dma-crossbar: Fix possible race condition with dma_inuse | expand |
On Thu, Sep 21, 2017 at 02:35:32PM +0300, Peter Ujfalusi wrote: > When looking for unused xbar_out lane we should also protect the set_bit() > call with the same mutex to protect against concurrent threads picking the > same ID. Applied, thanks Just a note, somehow the patches from you only are getting mangled for me. Curiously, the ones in patchworks are fine and I used to apply that, not sure what mail servers at our work places are doing!! > > Fixes: ec9bfa1e1a796 ("dmaengine: ti-dma-crossbar: dra7: Use bitops instead of idr") > Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com> > Cc: stable@vger.kernel.org > --- > drivers/dma/ti-dma-crossbar.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/dma/ti-dma-crossbar.c b/drivers/dma/ti-dma-crossbar.c > index 10ef9d5d5a66..7df910e7c348 100644 > --- a/drivers/dma/ti-dma-crossbar.c > +++ b/drivers/dma/ti-dma-crossbar.c > @@ -262,13 +262,14 @@ static void *ti_dra7_xbar_route_allocate(struct of_phandle_args *dma_spec, > mutex_lock(&xbar->mutex); > map->xbar_out = find_first_zero_bit(xbar->dma_inuse, > xbar->dma_requests); > - mutex_unlock(&xbar->mutex); > if (map->xbar_out == xbar->dma_requests) { > + mutex_unlock(&xbar->mutex); > dev_err(&pdev->dev, "Run out of free DMA requests\n"); > kfree(map); > return ERR_PTR(-ENOMEM); > } > set_bit(map->xbar_out, xbar->dma_inuse); > + mutex_unlock(&xbar->mutex); > > map->xbar_in = (u16)dma_spec->args[0]; > > -- > 2.14.1 > > > Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki > >
Vinod, Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki On 2017-09-21 20:30, Vinod Koul wrote: > On Thu, Sep 21, 2017 at 02:35:32PM +0300, Peter Ujfalusi wrote: >> When looking for unused xbar_out lane we should also protect the set_bit() >> call with the same mutex to protect against concurrent threads picking the >> same ID. > > Applied, thanks > > Just a note, somehow the patches from you only are getting mangled for me. > Curiously, the ones in patchworks are fine and I used to apply that, not sure > what mail servers at our work places are doing!! I'm afraid it is our mail servers where the mangling is happening :( We are trying to resolve this since early August when it first started to happen (and only for the guys in Finland). We are as frustrated as one can be, believe me. PS: And the mangler will add a footer on top of this mail - Péter
On Fri, Sep 22, 2017 at 10:03:59AM +0300, Peter Ujfalusi wrote: > Vinod, > > > Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki > > On 2017-09-21 20:30, Vinod Koul wrote: > > On Thu, Sep 21, 2017 at 02:35:32PM +0300, Peter Ujfalusi wrote: > >> When looking for unused xbar_out lane we should also protect the set_bit() > >> call with the same mutex to protect against concurrent threads picking the > >> same ID. > > > > Applied, thanks > > > > Just a note, somehow the patches from you only are getting mangled for me. > > Curiously, the ones in patchworks are fine and I used to apply that, not sure > > what mail servers at our work places are doing!! > > I'm afraid it is our mail servers where the mangling is happening :( > We are trying to resolve this since early August when it first started > to happen (and only for the guys in Finland). > We are as frustrated as one can be, believe me. > > PS: And the mangler will add a footer on top of this mail ah okay, consider sending using some other server then, or just put up a linux box in the corner like we do (@linux.intel.com)
diff --git a/drivers/dma/ti-dma-crossbar.c b/drivers/dma/ti-dma-crossbar.c index 10ef9d5d5a66..7df910e7c348 100644 --- a/drivers/dma/ti-dma-crossbar.c +++ b/drivers/dma/ti-dma-crossbar.c @@ -262,13 +262,14 @@ static void *ti_dra7_xbar_route_allocate(struct of_phandle_args *dma_spec, mutex_lock(&xbar->mutex); map->xbar_out = find_first_zero_bit(xbar->dma_inuse, xbar->dma_requests); - mutex_unlock(&xbar->mutex); if (map->xbar_out == xbar->dma_requests) { + mutex_unlock(&xbar->mutex); dev_err(&pdev->dev, "Run out of free DMA requests\n"); kfree(map); return ERR_PTR(-ENOMEM); } set_bit(map->xbar_out, xbar->dma_inuse); + mutex_unlock(&xbar->mutex); map->xbar_in = (u16)dma_spec->args[0];
When looking for unused xbar_out lane we should also protect the set_bit() call with the same mutex to protect against concurrent threads picking the same ID. Fixes: ec9bfa1e1a796 ("dmaengine: ti-dma-crossbar: dra7: Use bitops instead of idr") Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com> Cc: stable@vger.kernel.org --- drivers/dma/ti-dma-crossbar.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)