Just stumbled upon this trivial little patch that looks to have gotten lost in the shuffle. Seems it even got a reviewed-by from Brian [1]. So, uh I guess after almost 3 years...ping? -Tyrel [1] https://yhbt.net/lore/all/fd33df0e-012b-e437-c6e9-29cd0883808d@xxxxxxxxxxxxxxxxxx/ On 01/09/2019 08:59 PM, Tyrel Datwyler wrote: > During driver probe we allocate a dma region for our event pool. > Currently, zero is passed for the gfp_flags parameter. Driver probe > callbacks run in process context and we hold no locks so we can sleep > here if necessary. > > Fix by passing GFP_KERNEL explicitly to dma_alloc_coherent(). > > Signed-off-by: Tyrel Datwyler <tyreld@xxxxxxxxxxxxxxxxxx> > --- > drivers/scsi/ibmvscsi/ibmvscsi.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/scsi/ibmvscsi/ibmvscsi.c b/drivers/scsi/ibmvscsi/ibmvscsi.c > index cb8535e..10d5e77 100644 > --- a/drivers/scsi/ibmvscsi/ibmvscsi.c > +++ b/drivers/scsi/ibmvscsi/ibmvscsi.c > @@ -465,7 +465,7 @@ static int initialize_event_pool(struct event_pool *pool, > pool->iu_storage = > dma_alloc_coherent(hostdata->dev, > pool->size * sizeof(*pool->iu_storage), > - &pool->iu_token, 0); > + &pool->iu_token, GFP_KERNEL); > if (!pool->iu_storage) { > kfree(pool->events); > return -ENOMEM; >
