[Mesa-dev] [PATCH] clover: Fix host access validation for sub-buffer creation

[Francisco Jerez]

Aaron Watry <awatry at gmail.com> writes:

> On Mon, Apr 16, 2018, 5:24 PM Francisco Jerez <currojerez at riseup.net> wrote:
>
>> Aaron Watry <awatry at gmail.com> writes:
>>
>> >   From CL 1.2 Section 5.2.1:
>> >     CL_INVALID_VALUE if buffer was created with CL_MEM_HOST_WRITE_ONLY
>> and
>> >     flags specify CL_MEM_HOST_READ_ONLY , or if buffer was created with
>> >     CL_MEM_HOST_READ_ONLY and flags specify CL_MEM_HOST_WRITE_ONLY , or
>> if
>> >     buffer was created with CL_MEM_HOST_NO_ACCESS and flags specify
>> >     CL_MEM_HOST_READ_ONLY or CL_MEM_HOST_WRITE_ONLY .
>> >
>> > Fixes CL 1.2 CTS test/api get_buffer_info
>> >
>>
>> What combination of flags is the test-case providing for both the
>> parent and sub buffer?
>>
>
> The original motivation for this was a CTS test that was creating a sub
> buffer with flags of:
> CL_MEM_HOST_NO_ACCESS | CL_MEM_READ_WRITE
>
> With a parent buffer created as:
> CL_MEM_HOST_READ_ONLY | CL_MEM_READ_WRITE
>
> Which according to my reading of the spec should be allowed.
>

Right, I see.

>>
>> > Signed-off-by: Aaron Watry <awatry at gmail.com>
>> > Cc: Francisco Jerez <currojerez at riseup.net>
>> > ---
>> >  src/gallium/state_trackers/clover/api/memory.cpp | 8 ++++++--
>> >  1 file changed, 6 insertions(+), 2 deletions(-)
>> >
>> > diff --git a/src/gallium/state_trackers/clover/api/memory.cpp
>> b/src/gallium/state_trackers/clover/api/memory.cpp
>> > index 9b3cd8b1f5..451e8a8c56 100644
>> > --- a/src/gallium/state_trackers/clover/api/memory.cpp
>> > +++ b/src/gallium/state_trackers/clover/api/memory.cpp
>> > @@ -57,10 +57,14 @@ namespace {
>> >                                        parent.flags() &
>> host_access_flags) |
>> >                                       (parent.flags() & host_ptr_flags));
>> >
>> > -         if (~flags & parent.flags() &
>> > -             ((dev_access_flags & ~CL_MEM_READ_WRITE) |
>> host_access_flags))
>> > +         if (~flags & parent.flags() & (dev_access_flags &
>> ~CL_MEM_READ_WRITE))
>> >              throw error(CL_INVALID_VALUE);
>> >

I think you want to keep the hunk above and then do something along the
lines of:

+         if (!(flags & CL_MEM_HOST_NO_ACCESS) &&
+             (~flags & parent.flags() & host_access_flags))
+            throw error(CL_INVALID_VALUE);

>> > +         //Check if new host access flags cause a mismatch between
>> host-read/write-only.
>> > +         const cl_mem_flags new_flags = flags & ~(parent.flags()) &
>> ~CL_MEM_HOST_NO_ACCESS;
>> > +         if (new_flags & host_access_flags & parent.flags())
>> > +            throw error (CL_INVALID_VALUE);
>> > +
>>
>> This doesn't look correct to me, the condition will always evaluate to
>> zero, you're calculating the conjunction of ~parent.flags() and
>> parent.flags() which is zero, so the error will never be emitted.
>>
>
> I'll see what I can do. I agree with a fresh reading that it looks fishy at
> best.
>
> --Aaron
>
>>
>> >           return flags;
>> >
>> >        } else {
>> > --
>> > 2.14.1
>>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 227 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20180416/2b992991/attachment.sig>