[Bug 99419] Crash si_shader_select in Master Of Orion

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=99419

            Bug ID: 99419
           Summary: Crash si_shader_select in Master Of Orion
           Product: Mesa
           Version: git
          Hardware: Other
                OS: All
            Status: NEW
          Severity: normal
          Priority: medium
         Component: Drivers/Gallium/radeonsi
          Assignee: dri-devel at lists.freedesktop.org
          Reporter: andreas.schultes at googlemail.com
        QA Contact: dri-devel at lists.freedesktop.org

Created attachment 128966
  --> https://bugs.freedesktop.org/attachment.cgi?id=128966&action=edit
Backtrace

To reproduce crash click on a dropdown menu in settings. 

Crash caused by passing a invalid LLVMTargetMachineRef object in
si_shader_selector. Comment in si_shader.h says that should be only used in
si_init_shader_selector_async but that is not the case. It is also used in
si_build_shader_variant which called here also(see backtrace).

memorydump of tm object (frame 10)

x/60b sctx->tm
0x6c5b910:      224     219     246     103     41      127     0       0
0x6c5b918:      192     111     254     103     41      127     0       0
0x6c5b920:      0       0       0       0       0       0       0       0
0x6c5b928:      0       0       0       0       0       0       0       0
0x6c5b930:      72      185     197     6       0       0       0       0
0x6c5b938:      74      185     197     6       0       0       0       0
0x6c5b940:      80      185     197     6       0       0       0       0
0x6c5b948:      32      64      0       0

x/60b state->cso->tm
0x6c63ac0:      0       0       0       0       0       0       0       0
0x6c63ac8:      0       0       0       0       0       0       0       128
0x6c63ad0:      0       0       0       0       0       0       0       0
0x6c63ad8:      0       0       0       0       0       0       0       0
0x6c63ae0:      0       0       0       0       0       0       0       0
0x6c63ae8:      0       0       0       0       0       0       0       128
0x6c63af0:      0       0       0       0       0       0       0       0
0x6c63af8:      0       0       0       0

Adding state->cso->tm = sctx->tm; in si_shader_select prevents the crash. I
don't think that also fix the bug. There is still an incorrect rendering.

//disam shader
s_load_dwordx4 s[4:7], s[10:11], 0x0                ; C00A0105 00000000
s_load_dwordx4 s[0:3], s[2:3], 0x0                  ; C00A0001 00000000
v_mov_b32_e32 v0, 1.0                               ; 7E0002F2
s_waitcnt lgkmcnt(0)                                ; BF8C007F
buffer_load_format_xyzw v[3:6], v4, s[4:7], 0 idxen ; E00C2000 80010304
s_buffer_load_dword s5, s[0:3], 0x4                 ; C0220140 00000004
s_buffer_load_dword s9, s[0:3], 0x14                ; C0220240 00000014
s_buffer_load_dword s4, s[0:3], 0x0                 ; C0220100 00000000
s_buffer_load_dword s8, s[0:3], 0x10                ; C0220200 00000010
s_buffer_load_dword s6, s[0:3], 0x8                 ; C0220180 00000008
s_buffer_load_dword s10, s[0:3], 0x18               ; C0220280 00000018
s_buffer_load_dword s7, s[0:3], 0xc                 ; C02201C0 0000000C
s_buffer_load_dword s0, s[0:3], 0x1c                ; C0220000 0000001C
s_waitcnt vmcnt(0) lgkmcnt(0)                       ; BF8C0070
v_mul_f32_e32 v1, s5, v4                            ; 0A020805
v_mul_f32_e32 v4, s9, v4                            ; 0A080809
v_mac_f32_e32 v1, s4, v3                            ; 2C020604
v_mac_f32_e32 v4, s8, v3                            ; 2C080608
v_mac_f32_e32 v1, s6, v5                            ; 2C020A06
v_mac_f32_e32 v4, s10, v5                           ; 2C080A0A
v_mac_f32_e32 v1, s7, v6                            ; 2C020C07
v_mac_f32_e32 v4, s0, v6                            ; 2C080C00
v_mov_b32_e32 v3, 0                                 ; 7E060280
exp pos0 v1, v4, v3, v0 done                        ; C40008CF 00030401
s_waitcnt expcnt(0)                                 ; BF8C0F0F

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/dri-devel/attachments/20170115/c715e55b/attachment.html>