[PATCH v5 9/9] drm/xe: Implement capture of HWSP and HWCTX

[Souza, Jose]

On Wed, 2024-02-21 at 17:56 +0000, Souza, Jose wrote:
> On Wed, 2024-02-21 at 14:30 +0100, Maarten Lankhorst wrote:
> > Capture both the HWSP and the HW ctx page to allow easier replaying of
> > error dumps for debugging.
> > 
> > Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> > ---
> >  drivers/gpu/drm/xe/xe_lrc.c | 60 ++++++++++++++++++++++++++++++++++++-
> >  1 file changed, 59 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_lrc.c b/drivers/gpu/drm/xe/xe_lrc.c
> > index 20f235cdaca34..93c139f8d4fed 100644
> > --- a/drivers/gpu/drm/xe/xe_lrc.c
> > +++ b/drivers/gpu/drm/xe/xe_lrc.c
> > @@ -5,6 +5,8 @@
> >  
> >  #include "xe_lrc.h"
> >  
> > +#include <linux/ascii85.h>
> > +
> >  #include "instructions/xe_mi_commands.h"
> >  #include "instructions/xe_gfxpipe_commands.h"
> >  #include "regs/xe_engine_regs.h"
> > @@ -32,6 +34,10 @@
> >  #define ENGINE_INSTANCE_SHIFT			48
> >  
> >  struct xe_lrc_snapshot {
> > +	struct xe_bo *lrc_bo;
> > +	void *lrc_snapshot;
> > +	unsigned long lrc_size, lrc_offset;
> > +
> >  	u32 context_desc;
> >  	u32 head;
> >  	struct {
> > @@ -1333,19 +1339,43 @@ struct xe_lrc_snapshot *xe_lrc_snapshot_capture(struct xe_lrc *lrc)
> >  	snapshot->tail.memory = xe_lrc_read_ctx_reg(lrc, CTX_RING_TAIL);
> >  	snapshot->start_seqno = xe_lrc_start_seqno(lrc);
> >  	snapshot->seqno = xe_lrc_seqno(lrc);
> > +	snapshot->lrc_bo = xe_bo_get(lrc->bo);
> > +	snapshot->lrc_offset = xe_lrc_pphwsp_offset(lrc);
> > +	snapshot->lrc_size = lrc->bo->size - snapshot->lrc_offset;
> > +	snapshot->lrc_snapshot = NULL;
> >  	return snapshot;
> >  }
> >  
> >  void xe_lrc_snapshot_capture_delayed(struct xe_lrc_snapshot *snapshot)
> >  {
> > +	struct xe_bo *bo;
> > +	struct iosys_map src;
> > +
> >  	if (!snapshot)
> >  		return;
> >  
> > -	/* TODO: Copy status page */
> > +	bo = snapshot->lrc_bo;
> > +	snapshot->lrc_snapshot = kvmalloc(snapshot->lrc_size, GFP_KERNEL);
> > +	if (!snapshot->lrc_snapshot)
> > +		return;
> > +
> > +	dma_resv_lock(bo->ttm.base.resv, NULL);
> > +	if (!ttm_bo_vmap(&bo->ttm, &src)) {
> > +		xe_map_memcpy_from(xe_bo_device(bo),
> > +				   snapshot->lrc_snapshot, &src, snapshot->lrc_offset,
> > +				   snapshot->lrc_size);
> > +		ttm_bo_vunmap(&bo->ttm, &src);
> > +	} else {
> > +		kvfree(snapshot->lrc_snapshot);
> > +		snapshot->lrc_snapshot = NULL;
> > +	}
> > +	dma_resv_unlock(bo->ttm.base.resv);
> >  }
> >  
> >  void xe_lrc_snapshot_print(struct xe_lrc_snapshot *snapshot, struct drm_printer *p)
> >  {
> > +	unsigned long i;
> > +
> >  	if (!snapshot)
> >  		return;
> >  
> > @@ -1355,9 +1385,37 @@ void xe_lrc_snapshot_print(struct xe_lrc_snapshot *snapshot, struct drm_printer
> >  		   snapshot->tail.internal, snapshot->tail.memory);
> >  	drm_printf(p, "\tStart seqno: (memory) %d\n", snapshot->start_seqno);
> >  	drm_printf(p, "\tSeqno: (memory) %d\n", snapshot->seqno);
> > +
> > +	if (!snapshot->lrc_snapshot)
> > +		return;
> > +
> > +	drm_printf(p, "\t HWSP length: 0x%x\n", LRC_PPHWSP_SIZE);
> > +	drm_puts(p, "\tHWSP data: ");
> 
> drm_printf(p, "\t[HWSP].length: 0x%x\n", LRC_PPHWSP_SIZE);
> drm_puts(p, "\t[HWSP].data: ");
> 
> to make it consistent with VM state

HWSP and HWCTX should be placed in 'HW Engines' section under each engine?
What if a future HW moves out of GuC?

> 
> 
> > +
> > +	for (i = 0; i < LRC_PPHWSP_SIZE; i += sizeof(u32)) {
> > +		u32 *val = snapshot->lrc_snapshot + i;
> > +		char dumped[ASCII85_BUFSZ];
> > +
> > +		drm_puts(p, ascii85_encode(*val, dumped));
> > +	}
> 
> new line here
> 
> > +	drm_printf(p, "\n\tHWCTX length: 0x%lx\n", snapshot->lrc_size - LRC_PPHWSP_SIZE);
> > +	drm_puts(p, "\tHWCTX data: ");
> 
> drm_printf(p, "\n\t[HWCTX].length: 0x%lx\n", snapshot->lrc_size - LRC_PPHWSP_SIZE);
> drm_puts(p, "\t[HWCTX].data: ");
> 
> > +	for (; i < snapshot->lrc_size; i += sizeof(u32)) {
> > +		u32 *val = snapshot->lrc_snapshot + i;
> > +		char dumped[ASCII85_BUFSZ];
> > +
> > +		drm_puts(p, ascii85_encode(*val, dumped));
> > +	}
> > +	drm_puts(p, "\n");
> >  }
> >  
> >  void xe_lrc_snapshot_free(struct xe_lrc_snapshot *snapshot)
> >  {
> > +	if (!snapshot)
> > +		return;
> > +
> > +	kvfree(snapshot->lrc_snapshot);
> > +	if (snapshot->lrc_bo)
> > +		xe_bo_put(snapshot->lrc_bo);
> >  	kfree(snapshot);
> >  }
>