| Message ID | 20190509173849.11825-4-hch@lst.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/3] au1100fb: fix DMA API abuse | expand |
On 5/9/19 7:38 PM, Christoph Hellwig wrote: > Virtual addresses return from dma(m)_alloc_coherent are opaque in what > backs then, and drivers must not poke into them. Switch the driver > to use the generic DMA API mmap helper to avoid these problems. Change itself looks fine but the patch description doesn't match what the patch is actually doing (there is no conversion to DMA API helper because the driver is already using it). Please correct it. Thank you. > Signed-off-by: Christoph Hellwig <hch@lst.de> > --- > drivers/video/fbdev/jz4740_fb.c | 7 ------- > 1 file changed, 7 deletions(-) > diff --git a/drivers/video/fbdev/jz4740_fb.c b/drivers/video/fbdev/jz4740_fb.c > index b57df83fdbd3..b95cdfa9e0a1 100644 > --- a/drivers/video/fbdev/jz4740_fb.c > +++ b/drivers/video/fbdev/jz4740_fb.c > @@ -466,7 +466,6 @@ static int jzfb_alloc_devmem(struct jzfb *jzfb) > { > int max_videosize = 0; > struct fb_videomode *mode = jzfb->pdata->modes; > - void *page; > int i; > > for (i = 0; i < jzfb->pdata->num_modes; ++mode, ++i) { > @@ -491,12 +490,6 @@ static int jzfb_alloc_devmem(struct jzfb *jzfb) > if (!jzfb->vidmem) > goto err_free_framedesc; > > - for (page = jzfb->vidmem; > - page < jzfb->vidmem + PAGE_ALIGN(jzfb->vidmem_size); > - page += PAGE_SIZE) { > - SetPageReserved(virt_to_page(page)); > - } > - > jzfb->framedesc->next = jzfb->framedesc_phys; > jzfb->framedesc->addr = jzfb->vidmem_phys; > jzfb->framedesc->id = 0xdeafbead; Best regards, -- Bartlomiej Zolnierkiewicz Samsung R&D Institute Poland Samsung Electronics
Hi, On 6/7/19 2:21 PM, Bartlomiej Zolnierkiewicz wrote: > > On 5/9/19 7:38 PM, Christoph Hellwig wrote: >> Virtual addresses return from dma(m)_alloc_coherent are opaque in what >> backs then, and drivers must not poke into them. Switch the driver >> to use the generic DMA API mmap helper to avoid these problems. > Change itself looks fine but the patch description doesn't match what > the patch is actually doing (there is no conversion to DMA API helper > because the driver is already using it). Please correct it. Thank you. I've just removed the "Switch the driver.." sentence myself from the patch description and applied the change, thanks! >> Signed-off-by: Christoph Hellwig <hch@lst.de> >> --- >> drivers/video/fbdev/jz4740_fb.c | 7 ------- >> 1 file changed, 7 deletions(-) >> diff --git a/drivers/video/fbdev/jz4740_fb.c b/drivers/video/fbdev/jz4740_fb.c >> index b57df83fdbd3..b95cdfa9e0a1 100644 >> --- a/drivers/video/fbdev/jz4740_fb.c >> +++ b/drivers/video/fbdev/jz4740_fb.c >> @@ -466,7 +466,6 @@ static int jzfb_alloc_devmem(struct jzfb *jzfb) >> { >> int max_videosize = 0; >> struct fb_videomode *mode = jzfb->pdata->modes; >> - void *page; >> int i; >> >> for (i = 0; i < jzfb->pdata->num_modes; ++mode, ++i) { >> @@ -491,12 +490,6 @@ static int jzfb_alloc_devmem(struct jzfb *jzfb) >> if (!jzfb->vidmem) >> goto err_free_framedesc; >> >> - for (page = jzfb->vidmem; >> - page < jzfb->vidmem + PAGE_ALIGN(jzfb->vidmem_size); >> - page += PAGE_SIZE) { >> - SetPageReserved(virt_to_page(page)); >> - } >> - >> jzfb->framedesc->next = jzfb->framedesc_phys; >> jzfb->framedesc->addr = jzfb->vidmem_phys; >> jzfb->framedesc->id = 0xdeafbead; Best regards, -- Bartlomiej Zolnierkiewicz Samsung R&D Institute Poland Samsung Electronics
On Fri, Jun 21, 2019 at 01:32:16PM +0200, Bartlomiej Zolnierkiewicz wrote: > I've just removed the "Switch the driver.." sentence myself from > the patch description and applied the change, thanks! Sorry for not getting back to you earlier and thanks for fixing this up and applying the patch.
diff --git a/drivers/video/fbdev/jz4740_fb.c b/drivers/video/fbdev/jz4740_fb.c index b57df83fdbd3..b95cdfa9e0a1 100644 --- a/drivers/video/fbdev/jz4740_fb.c +++ b/drivers/video/fbdev/jz4740_fb.c @@ -466,7 +466,6 @@ static int jzfb_alloc_devmem(struct jzfb *jzfb) { int max_videosize = 0; struct fb_videomode *mode = jzfb->pdata->modes; - void *page; int i; for (i = 0; i < jzfb->pdata->num_modes; ++mode, ++i) { @@ -491,12 +490,6 @@ static int jzfb_alloc_devmem(struct jzfb *jzfb) if (!jzfb->vidmem) goto err_free_framedesc; - for (page = jzfb->vidmem; - page < jzfb->vidmem + PAGE_ALIGN(jzfb->vidmem_size); - page += PAGE_SIZE) { - SetPageReserved(virt_to_page(page)); - } - jzfb->framedesc->next = jzfb->framedesc_phys; jzfb->framedesc->addr = jzfb->vidmem_phys; jzfb->framedesc->id = 0xdeafbead;
Virtual addresses return from dma(m)_alloc_coherent are opaque in what backs then, and drivers must not poke into them. Switch the driver to use the generic DMA API mmap helper to avoid these problems. Signed-off-by: Christoph Hellwig <hch@lst.de> --- drivers/video/fbdev/jz4740_fb.c | 7 ------- 1 file changed, 7 deletions(-)