On 09/01/18 13:37, Michel Thierry wrote:
On 09/01/18 12:46, Chris Wilson wrote:
Quoting Michal Wajdeczko (2018-01-09 20:39:09)
On Tue, 09 Jan 2018 20:33:55 +0100, Michel Thierry
<michel.thierry@xxxxxxxxx> wrote:
Instead of using local string names that we will have to keep
maintaining, use the engine->name directly.
Suggested-by: Michal Wajdeczko <michal.wajdeczko@xxxxxxxxx>
Signed-off-by: Michel Thierry <michel.thierry@xxxxxxxxx>
Cc: Michal Wajdeczko <michal.wajdeczko@xxxxxxxxx>
I'd a patch to do this, I just had to make sure the tooling was ready
for a name change. At the time, changing this string broke a few igt and
intel_error_decode.
I think your igt patch "lib: Fix up internal engine names (again)"
covered it.
Not sure if there are more tools looking at the engines' registers in
the error_state outside igt.
---
drivers/gpu/drm/i915/i915_gpu_error.c | 23 ++++++++++-------------
1 file changed, 10 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c
b/drivers/gpu/drm/i915/i915_gpu_error.c
index 94499c24f279..db95ecacdace 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -34,16 +34,12 @@
#include "i915_drv.h"
-static const char *engine_str(int engine)
-{
- switch (engine) {
- case RCS: return "render";
- case VCS: return "bsd";
- case BCS: return "blt";
- case VECS: return "vebox";
- case VCS2: return "bsd2";
- default: return "";
- }
+static const char *engine_str(struct drm_i915_private *i915, int
engine_id)
+{
+ if (!i915->engine[engine_id])
+ return "";
While unlikely, empty string may be misleading, so maybe better we
return
"<invalid>" or at least "?" here. Also maybe we can do as part of more
global helper function like
static inline const char* intel_engine_name(struct intel_engine_cs
*engine)
{
return engine ? engine->name : "<invalid>";
}
static inline struct intel_engine_cs *
intel_engine_lookup(struct drm_i915_private *i915, int engine_id)
{
if (engine_id < 0 || engine_id > I915_MAX_ENGINES)
return NULL;
return i915->engine[engine_id];
}
and then
static const char *engine_str(struct drm_i915_private *i915, int
engine_id)
{
return intel_engine_name(intel_engine_lookup(i915, engine_id));
}
No need, this is all internal, perma-allocated structs.
i915_gpu_info_open can pass engine_id=-1, better have some protection.
E.g. cat /sys/kernel/debug/dri/0/i915_gpu_info (while idle)
[ 84.448504] GEM_BUG_ON(engine_id == -1)
[ 84.448591] ------------[ cut here ]------------
[ 84.448630] kernel BUG at drivers/gpu/drm/i915/i915_gpu_error.c:39!
[ 84.448673] invalid opcode: 0000 [#1] SMP
[ 84.448696] Modules linked in: rfcomm bnep nls_iso8859_1
...
[ 84.449425] Call Trace:
[ 84.449463] print_error_buffers+0x166/0x210 [i915]
[ 84.449510] i915_error_state_to_str+0x923/0x1210 [i915]
[ 84.449541] ? create_object+0x23f/0x2e0
[ 84.449584] ? i915_error_state_buf_init+0x59/0xc0 [i915]
[ 84.449613] ? rcu_read_lock_sched_held+0x49/0x90
[ 84.449644] ? kmalloc_order_trace+0x10d/0x170
[ 84.449685] gpu_state_read+0x56/0xb0 [i915]
[ 84.449712] full_proxy_read+0x4d/0x70
[ 84.449736] __vfs_read+0x23/0x120
[ 84.449759] vfs_read+0xa0/0x140
[ 84.449779] SyS_read+0x45/0xa0
[ 84.449801] entry_SYSCALL_64_fastpath+0x23/0x9a
[ 84.449834] RIP: 0033:0x7f2a2d3d4230
[ 84.449856] RSP: 002b:00007ffda36960e8 EFLAGS: 00000246
[ 84.449861] Code: 15 83 fe 04 7f 27 48 63 f6 48 8b 84 f7 08 46 00 00
48 83 c0 08 c3 48 c7 c6 e3 03 59 c0 48 c7 c7 f3 03 59 c0 31 c0 e8 66 75
b9 c6 <0f> 0b 48 c7 c6 05 04 59 c0 48 c7 c7 f3 03 59 c0 31 c0 e8 4f 75
[ 84.450028] RIP: engine_str+0x34/0x50 [i915] RSP: ffffb7e142b53c98
[ 89.410577] ---[ end trace 5eacc9ca2b6d1e5d ]---
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
