[Bug 105301] The big SKQP bug

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=105301

--- Comment #24 from Jianxun Zhang <jianxun.zhang at intel.com> ---
(In reply to Carlos Santa from comment #10)
> Hi,
> 
> Here are more details to the CTS issue we are trying to solve.
> 
> We are able to repro 100% in the Android container under Chrome OS running
> Linux kernel 4.4 and Mesa 18. We can also repro under Celadon/NUC/drm tip
> running kernel 4.19 and Mesa 18.
> 
> The code from the upper layer is here :
> http://androidxref.com/9.0.0_r3/xref/external/skia/gm/mixedtextblobs.cpp#20
> 
> There are 3 main draw calls()
> 
> 26    SkPaint paint(skPaint);
> 28    canvas->drawRect(clipRect, clipHairline);
> 30    canvas->drawTextBlob(blob, 0, 0, paint);
> 33    canvas->drawTextBlob(blob, 0, 0, paint);
> 
> We are still trying to see what these functions are in GL apis (more soon).
> 
> The IPHER, which points to the command that hung is at IPEHR: 0x78260000,
> and that address point to 3DSTATE_BINDING_TABLE_POINTERS_VS
>  
> 0xe60003ac:      0x784f0000: 3D UNKNOWN: 3d_965 opcode = 0x784f
> 0xe60003b0:      0x90000100: UNKNOWN
> 0xe60003b4:      0x78230000: 3D UNKNOWN: 3d_965 opcode = 0x7823
> 0xe60003b8:      0x00000140: MI_NOOP
> 0xe60003bc:      0x78260000: 3DSTATE_BINDING_TABLE_POINTERS_VS
> 0xe60003c0:      0x00000000:    dword 1
> 0xe60003c4:      0x78270000: 3DSTATE_BINDING_TABLE_POINTERS_HS
> 0xe60003c8:      0x00000000:    dword 1
> 0xe60003cc:      0x78280000: 3DSTATE_BINDING_TABLE_POINTERS_DS
> 0xe60003d0:      0x00000000:    dword 1
> 0xe60003d4:      0x78290000: 3DSTATE_BINDING_TABLE_POINTERS_GS
> 0xe60003d8:      0x00000000:    dword 1
> 0xe60003dc:      0x782a0000: 3DSTATE_BINDING_TABLE_POINTERS_PS
> 0xe60003e0:      0x00000160:    dword 1
>  
> The documentation says this: The 3DSTATE_BINDING_TABLE_POINTERS_VS command
> is used to define the location of fixed functions'
> BINDING_TABLE_STATE. Only some of the fixed functions utilize binding tables.

The progress could pass this stage but just want to keep the record straight.

I believe this is from the crash dump provided in the following comment #12. I
think the correct address is another location in the batch buffer:

0xe60034bc:      0x00000000: MI_NOOP
0xe60034c0:      0x00000000: MI_NOOP
0xe60034c4:      0xd00103c0: UNKNOWN
0xe60034c8:      0xffffffff: UNKNOWN
0xe60034cc:      0x78260000: 3DSTATE_BINDING_TABLE_POINTERS_VS
0xe60034d0:      0x00000000:    dword 1
0xe60034d4:      0x782a0000: 3DSTATE_BINDING_TABLE_POINTERS_PS

Correct me if I am wrong.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20181030/5de2e988/attachment.html>