Re: [PATCH v2 2/2] drm/etnaviv: print MMU exception cause

On Fr, 2019-09-13 at 07:50 +0200, Christian Gmeiner wrote:

Might be useful when debugging MMU exceptions.

Changes in V2:

• Use a static array of string for error message as suggested by Lucas Stach.

Please move those changelogs below the 3 dashes, so they don't end up in the commit message. They don't really add any value to the persistent kernel history.

Signed-off-by: Christian Gmeiner christian.gmeiner@gmail.com

drivers/gpu/drm/etnaviv/etnaviv_gpu.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+)

diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c index d47d1a8e0219..b8cd85153fe0 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c @@ -1351,6 +1351,15 @@ static void sync_point_worker(struct work_struct *work)

static void dump_mmu_fault(struct etnaviv_gpu *gpu) {

• static const char *errors[] = {

• "slave not present",
  

• "page not present",
  

• "write violation",
  

• "out of bound",
  

• "read security violation",
  

• "write security violation",
  

• };
• u32 status_reg, status; int i;

@@ -1364,10 +1373,16 @@ static void dump_mmu_fault(struct etnaviv_gpu *gpu)

for (i = 0; i < 4; i++) { u32 address_reg;

• const char *error = "unknown state";
  
  

  if (!(status & (VIVS_MMUv2_STATUS_EXCEPTION0__MASK << (i * 4)))) continue;

• if (status < ARRAY_SIZE(errors))
  

• 	error = errors[status];
  

Huh? This won't work. The status register is a bitfield, not an integer so you need to map the bit position to the array index via ffs() or something like that.

Regards, Lucas

• dev_err_ratelimited(gpu->dev, "MMU %d %s\n", i, error);
  

• if (gpu->sec_mode == ETNA_SEC_NONE) address_reg = VIVS_MMUv2_EXCEPTION_ADDR(i); else