Message ID | alpine.SOC.1.00.1310101444090.15296@math.ut.ee (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Thu, Oct 10, 2013 at 7:51 AM, Meelis Roos <mroos@linux.ee> wrote: > To prevent hangs on non-PC machines (e.g. sparc64), probe Radeon ROM > from ATI IGP only on X86. Fixes hang in this place and allows PCI radeon > detection to move on to next problem. NACK. All this function does it attempt to read the rom from the framebuffer PCI BAR. If you get hangs reading the BAR, then you have larger problems on your platform. Also, there are non-x86 platforms that IGP chips which this may break. Alex > > Signed-off-by: Meelis Roos <mroos@linux.ee> > > diff --git a/drivers/gpu/drm/radeon/radeon_bios.c b/drivers/gpu/drm/radeon/radeon_bios.c > index 061b227..48ef97c 100644 > --- a/drivers/gpu/drm/radeon/radeon_bios.c > +++ b/drivers/gpu/drm/radeon/radeon_bios.c > @@ -49,6 +49,11 @@ static bool igp_read_bios_from_vram(struct radeon_device *rdev) > resource_size_t vram_base; > resource_size_t size = 256 * 1024; /* ??? */ > > +#ifndef CONFIG_X86 > + /* IGP only exists on X86 (32- and 64-bit) */ > + return false; > +#endif > + > if (!(rdev->flags & RADEON_IS_IGP)) > if (!radeon_card_posted(rdev)) > return false; > > -- > Meelis Roos (mroos@linux.ee)
> On Thu, Oct 10, 2013 at 7:51 AM, Meelis Roos <mroos@linux.ee> wrote: > > To prevent hangs on non-PC machines (e.g. sparc64), probe Radeon ROM > > from ATI IGP only on X86. Fixes hang in this place and allows PCI radeon > > detection to move on to next problem. > > NACK. All this function does it attempt to read the rom from the > framebuffer PCI BAR. If you get hangs reading the BAR, then you have > larger problems on your platform. Also, there are non-x86 platforms > that IGP chips which this may break. OK. But the code seems to just dereference addresses returned from ioremap, but to the best of my knowledge, these are not universally (like on each arch) safe to read without readb/readw/readl. Should I convert these to readb for test? > > Signed-off-by: Meelis Roos <mroos@linux.ee> > > > > diff --git a/drivers/gpu/drm/radeon/radeon_bios.c b/drivers/gpu/drm/radeon/radeon_bios.c > > index 061b227..48ef97c 100644 > > --- a/drivers/gpu/drm/radeon/radeon_bios.c > > +++ b/drivers/gpu/drm/radeon/radeon_bios.c > > @@ -49,6 +49,11 @@ static bool igp_read_bios_from_vram(struct radeon_device *rdev) > > resource_size_t vram_base; > > resource_size_t size = 256 * 1024; /* ??? */ > > > > +#ifndef CONFIG_X86 > > + /* IGP only exists on X86 (32- and 64-bit) */ > > + return false; > > +#endif > > + > > if (!(rdev->flags & RADEON_IS_IGP)) > > if (!radeon_card_posted(rdev)) > > return false; > > > > -- > > Meelis Roos (mroos@linux.ee) >
From: Meelis Roos <mroos@linux.ee> Date: Thu, 10 Oct 2013 18:26:39 +0300 (EEST) >> On Thu, Oct 10, 2013 at 7:51 AM, Meelis Roos <mroos@linux.ee> wrote: >> > To prevent hangs on non-PC machines (e.g. sparc64), probe Radeon ROM >> > from ATI IGP only on X86. Fixes hang in this place and allows PCI radeon >> > detection to move on to next problem. >> >> NACK. All this function does it attempt to read the rom from the >> framebuffer PCI BAR. If you get hangs reading the BAR, then you have >> larger problems on your platform. Also, there are non-x86 platforms >> that IGP chips which this may break. > > OK. But the code seems to just dereference addresses returned from > ioremap, but to the best of my knowledge, these are not universally > (like on each arch) safe to read without readb/readw/readl. Should I > convert these to readb for test? That's correct, ioremap() returned pointers must not be directly dereferenced, and must be accessed through the appropriate accessors.
On Thu, Oct 10, 2013 at 11:26 AM, Meelis Roos <mroos@linux.ee> wrote: >> On Thu, Oct 10, 2013 at 7:51 AM, Meelis Roos <mroos@linux.ee> wrote: >> > To prevent hangs on non-PC machines (e.g. sparc64), probe Radeon ROM >> > from ATI IGP only on X86. Fixes hang in this place and allows PCI radeon >> > detection to move on to next problem. >> >> NACK. All this function does it attempt to read the rom from the >> framebuffer PCI BAR. If you get hangs reading the BAR, then you have >> larger problems on your platform. Also, there are non-x86 platforms >> that IGP chips which this may break. > > OK. But the code seems to just dereference addresses returned from > ioremap, but to the best of my knowledge, these are not universally > (like on each arch) safe to read without readb/readw/readl. Should I > convert these to readb for test? Yeah, that would be great. Thanks, Alex > >> > Signed-off-by: Meelis Roos <mroos@linux.ee> >> > >> > diff --git a/drivers/gpu/drm/radeon/radeon_bios.c b/drivers/gpu/drm/radeon/radeon_bios.c >> > index 061b227..48ef97c 100644 >> > --- a/drivers/gpu/drm/radeon/radeon_bios.c >> > +++ b/drivers/gpu/drm/radeon/radeon_bios.c >> > @@ -49,6 +49,11 @@ static bool igp_read_bios_from_vram(struct radeon_device *rdev) >> > resource_size_t vram_base; >> > resource_size_t size = 256 * 1024; /* ??? */ >> > >> > +#ifndef CONFIG_X86 >> > + /* IGP only exists on X86 (32- and 64-bit) */ >> > + return false; >> > +#endif >> > + >> > if (!(rdev->flags & RADEON_IS_IGP)) >> > if (!radeon_card_posted(rdev)) >> > return false; >> > >> > -- >> > Meelis Roos (mroos@linux.ee) >> > > -- > Meelis Roos (mroos@linux.ee)
diff --git a/drivers/gpu/drm/radeon/radeon_bios.c b/drivers/gpu/drm/radeon/radeon_bios.c index 061b227..48ef97c 100644 --- a/drivers/gpu/drm/radeon/radeon_bios.c +++ b/drivers/gpu/drm/radeon/radeon_bios.c @@ -49,6 +49,11 @@ static bool igp_read_bios_from_vram(struct radeon_device *rdev) resource_size_t vram_base; resource_size_t size = 256 * 1024; /* ??? */ +#ifndef CONFIG_X86 + /* IGP only exists on X86 (32- and 64-bit) */ + return false; +#endif + if (!(rdev->flags & RADEON_IS_IGP)) if (!radeon_card_posted(rdev)) return false;
To prevent hangs on non-PC machines (e.g. sparc64), probe Radeon ROM from ATI IGP only on X86. Fixes hang in this place and allows PCI radeon detection to move on to next problem. Signed-off-by: Meelis Roos <mroos@linux.ee>