drm/i915: Allow parsing of unsized batches
authorJon Bloomfield <jon.bloomfield@intel.com>
Wed, 1 Aug 2018 16:45:50 +0000 (09:45 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 12 Nov 2019 18:21:24 +0000 (19:21 +0100)
commit 435e8fc059dbe0eec823a75c22da2972390ba9e0 upstream.

In "drm/i915: Add support for mandatory cmdparsing" we introduced the
concept of mandatory parsing. This allows the cmdparser to be invoked
even when user passes batch_len=0 to the execbuf ioctl's.

However, the cmdparser needs to know the extents of the buffer being
scanned. Refactor the code to ensure the cmdparser uses the actual
object size, instead of the incoming length, if user passes 0.

Signed-off-by: Jon Bloomfield <jon.bloomfield@intel.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Takashi Iwai <tiwai@suse.de>
Cc: Tyler Hicks <tyhicks@canonical.com>
Reviewed-by: Chris Wilson <chris.p.wilson@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/gpu/drm/i915/i915_gem_execbuffer.c

index 3fc9acc..81559a1 100644 (file)
@@ -310,7 +310,8 @@ static inline u64 gen8_noncanonical_addr(u64 address)
 static inline bool eb_use_cmdparser(const struct i915_execbuffer *eb)
 {
        return intel_engine_requires_cmd_parser(eb->engine) ||
-               (intel_engine_using_cmd_parser(eb->engine) && eb->batch_len);
+               (intel_engine_using_cmd_parser(eb->engine) &&
+                eb->args->batch_len);
 }
 
 static int eb_create(struct i915_execbuffer *eb)
@@ -2341,6 +2342,9 @@ i915_gem_do_execbuffer(struct drm_device *dev,
                goto err_vma;
        }
 
+       if (eb.batch_len == 0)
+               eb.batch_len = eb.batch->size - eb.batch_start_offset;
+
        if (eb_use_cmdparser(&eb)) {
                struct i915_vma *vma;
 
@@ -2351,9 +2355,6 @@ i915_gem_do_execbuffer(struct drm_device *dev,
                }
        }
 
-       if (eb.batch_len == 0)
-               eb.batch_len = eb.batch->size - eb.batch_start_offset;
-
        /*
         * snb/ivb/vlv conflate the "batch in ppgtt" bit with the "non-secure
         * batch" bit. Hence we need to pin secure batches into the global gtt.