[PATCH 07/18] drm/i915/gem: Set the watchdog timeout as part of intel_context creation

[Jason Ekstrand]

On Fri, Apr 16, 2021 at 3:01 AM Tvrtko Ursulin
<tvrtko.ursulin at linux.intel.com> wrote:
>
>
> On 15/04/2021 17:40, Jason Ekstrand wrote:
> > Instead of handling it like a context param, we add a little wrapper for
> > intel_context_create() whenever it's used to create a user context.
> > This wrapper always sets the timeout for us.  This fixes a bug where a
> > client could get around the timeout by swapping out the set of engines
> > using SETPARAM.
>
> ctx_setparam -> set_engines -> intel_context_set_gem ->
> intel_context_set_watchdog_us applies the timeout so I don't see the bug.

Yup, you're right.  This isn't the first time I've missed that
intel_context_set_gem does something useful.  I've reworked this patch
to just move setting the watchdog there.  I think that should nicely
cover all cases.

--Jason

> Regards,
>
> Tvrtko
>
> > ---
> >   drivers/gpu/drm/i915/gem/i915_gem_context.c   | 58 ++++++-------------
> >   drivers/gpu/drm/i915/gt/intel_context_param.h |  3 +-
> >   2 files changed, 20 insertions(+), 41 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> > index 0b35ab9509cb2..857a60469d510 100644
> > --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
> > +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> > @@ -287,6 +287,23 @@ engines_notify(struct i915_sw_fence *fence, enum i915_sw_fence_notify state)
> >       return NOTIFY_DONE;
> >   }
> >
> > +static struct intel_context *
> > +intel_user_context_create(struct drm_i915_private *i915,
> > +                       struct intel_engine_cs *engine)
> > +{
> > +     struct intel_context *ce = intel_context_create(engine);
> > +     if (IS_ERR(ce))
> > +             return ce;
> > +
> > +     if (IS_ACTIVE(CONFIG_DRM_I915_REQUEST_TIMEOUT) &&
> > +         i915->params.request_timeout_ms) {
> > +             unsigned long timeout_us = i915->params.request_timeout_ms * 1000;
> > +             intel_context_set_watchdog_us(ce, (uintptr_t)timeout_us);
> > +     }
> > +
> > +     return ce;
> > +}
> > +
> >   static struct i915_gem_engines *alloc_engines(unsigned int count)
> >   {
> >       struct i915_gem_engines *e;
> > @@ -319,7 +336,7 @@ static struct i915_gem_engines *default_engines(struct i915_gem_context *ctx)
> >               GEM_BUG_ON(engine->legacy_idx >= I915_NUM_ENGINES);
> >               GEM_BUG_ON(e->engines[engine->legacy_idx]);
> >
> > -             ce = intel_context_create(engine);
> > +             ce = intel_user_context_create(ctx->i915, engine);
> >               if (IS_ERR(ce)) {
> >                       __free_engines(e, e->num_engines + 1);
> >                       return ERR_CAST(ce);
> > @@ -759,41 +776,6 @@ static void __assign_ppgtt(struct i915_gem_context *ctx,
> >               i915_vm_close(vm);
> >   }
> >
> > -static int __apply_watchdog(struct intel_context *ce, void *timeout_us)
> > -{
> > -     return intel_context_set_watchdog_us(ce, (uintptr_t)timeout_us);
> > -}
> > -
> > -static int
> > -__set_watchdog(struct i915_gem_context *ctx, unsigned long timeout_us)
> > -{
> > -     int ret;
> > -
> > -     ret = context_apply_all(ctx, __apply_watchdog,
> > -                             (void *)(uintptr_t)timeout_us);
> > -     if (!ret)
> > -             ctx->watchdog.timeout_us = timeout_us;
> > -
> > -     return ret;
> > -}
> > -
> > -static void __set_default_fence_expiry(struct i915_gem_context *ctx)
> > -{
> > -     struct drm_i915_private *i915 = ctx->i915;
> > -     int ret;
> > -
> > -     if (!IS_ACTIVE(CONFIG_DRM_I915_REQUEST_TIMEOUT) ||
> > -         !i915->params.request_timeout_ms)
> > -             return;
> > -
> > -     /* Default expiry for user fences. */
> > -     ret = __set_watchdog(ctx, i915->params.request_timeout_ms * 1000);
> > -     if (ret)
> > -             drm_notice(&i915->drm,
> > -                        "Failed to configure default fence expiry! (%d)",
> > -                        ret);
> > -}
> > -
> >   static struct i915_gem_context *
> >   i915_gem_create_context(struct drm_i915_private *i915, unsigned int flags)
> >   {
> > @@ -836,8 +818,6 @@ i915_gem_create_context(struct drm_i915_private *i915, unsigned int flags)
> >               }
> >       }
> >
> > -     __set_default_fence_expiry(ctx);
> > -
> >       trace_i915_context_create(ctx);
> >
> >       return ctx;
> > @@ -1715,7 +1695,7 @@ set_engines(struct i915_gem_context *ctx,
> >                       return -ENOENT;
> >               }
> >
> > -             ce = intel_context_create(engine);
> > +             ce = intel_user_context_create(i915, engine);
> >               if (IS_ERR(ce)) {
> >                       __free_engines(set.engines, n);
> >                       return PTR_ERR(ce);
> > diff --git a/drivers/gpu/drm/i915/gt/intel_context_param.h b/drivers/gpu/drm/i915/gt/intel_context_param.h
> > index dffedd983693d..0c69cb42d075c 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_context_param.h
> > +++ b/drivers/gpu/drm/i915/gt/intel_context_param.h
> > @@ -10,11 +10,10 @@
> >
> >   #include "intel_context.h"
> >
> > -static inline int
> > +static inline void
> >   intel_context_set_watchdog_us(struct intel_context *ce, u64 timeout_us)
> >   {
> >       ce->watchdog.timeout_us = timeout_us;
> > -     return 0;
> >   }
> >
> >   #endif /* INTEL_CONTEXT_PARAM_H */
> >