On Sun, Jan 11, 2026 at 12:54 AM Michael S. Tsirkin <[email protected]> wrote: > > On Fri, Jan 09, 2026 at 04:24:27PM +0100, Eugenio Pérez wrote: > > diff --git a/drivers/vdpa/vdpa_user/vduse_dev.c > > b/drivers/vdpa/vdpa_user/vduse_dev.c > > index 82ee476d45e0..675da1465e0e 100644 > > --- a/drivers/vdpa/vdpa_user/vduse_dev.c > > +++ b/drivers/vdpa/vdpa_user/vduse_dev.c > > @@ -923,16 +923,24 @@ static void *vduse_dev_alloc_coherent(union > > virtio_map token, size_t size, > > if (!token.group) > > return NULL; > > > > + addr = alloc_pages_exact(size, flag); > > + if (!addr) > > + return NULL; > > + > > So addr has allocated pages here ... > > > > vdev = token.group->dev; > > domain = vdev->domain; > > addr = vduse_domain_alloc_coherent(domain, size, > > - (dma_addr_t *)&iova, flag); > > + (dma_addr_t *)&iova, addr); > > and then is overwritten here ... > > > if (!addr) > > - return NULL; > > + goto err; > > except on error where we go to err ... > > > > > *dma_addr = (dma_addr_t)iova; > > > > return addr; > > + > > +err: > > + free_pages_exact(addr, size); > > only to try and free NULL. will leak the original pages, will it not. >
Right, I missed that in the conversion. I'll change the function vduse_domain_alloc_coherent so it returns dma_addr instead of orig, making all the code simpler. Thanks! > > + return NULL; > > } > > > > static void vduse_dev_free_coherent(union virtio_map token, size_t size, >
