drm/xe/guc: Fix out-of-bound while enabling engine activity stats
authorMichal Wajdeczko <michal.wajdeczko@intel.com>
Mon, 14 Apr 2025 20:23:46 +0000 (22:23 +0200)
committerMichal Wajdeczko <michal.wajdeczko@intel.com>
Thu, 17 Apr 2025 19:57:18 +0000 (21:57 +0200)
In the PF mode we allocate array of struct engine_activity_group
that holds activity data split for the PF and all potential VFs.
But while preparing data for use by VFs we ended with bad index.

 [ ] BUG: KASAN: slab-out-of-bounds in xe_guc_engine_activity_function_stats+0x41e/0x4f0 [xe]
 [ ] Call Trace:
 [ ]  <TASK>
 [ ]  dump_stack_lvl+0x91/0xf0
 [ ]  print_report+0xd1/0x680
 [ ]  ? __virt_addr_valid+0x23a/0x440
 [ ]  ? kasan_addr_to_slab+0xd/0xb0
 [ ]  kasan_report+0xe7/0x130
 [ ]  ? xe_guc_engine_activity_function_stats+0x41e/0x4f0 [xe]
 [ ]  ? xe_guc_engine_activity_function_stats+0x41e/0x4f0 [xe]
 [ ]  __asan_report_store8_noabort+0x17/0x30
 [ ]  xe_guc_engine_activity_function_stats+0x41e/0x4f0 [xe]
 [ ]  pf_engine_activity_stats+0x1b6/0x7f0 [xe]
 [ ]  ? kobject_put+0x5f/0x470
 [ ]  xe_pci_sriov_configure+0x28c9/0x3270 [xe]
 [ ]  ? __pfx_dev_attr_store+0x10/0x10
 [ ]  ? kstrtoull+0x3b/0x70
 [ ]  ? __pfx___lock_acquire+0x10/0x10
 [ ]  ? kstrtou16+0x65/0xf0
 [ ]  sriov_numvfs_store+0x20c/0x400
 [ ]  ? __pfx_sriov_numvfs_store+0x10/0x10
 [ ]  ? __pfx__copy_from_iter+0x10/0x10
 [ ]  ? __pfx_dev_attr_store+0x10/0x10
 [ ]  dev_attr_store+0x3b/0x80
 [ ]  ? sysfs_file_ops+0x135/0x190

Fixes: 2de3f38fbf89 ("drm/xe: Add support for per-function engine activity")
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Cc: Riana Tauro <riana.tauro@intel.com>
Reviewed-by: Riana Tauro <riana.tauro@intel.com>
Link: https://lore.kernel.org/r/20250414202347.1909-1-michal.wajdeczko@intel.com
drivers/gpu/drm/xe/xe_guc_engine_activity.c

index b96fea78df8b4b27b1684e042de806f2f1c77819..0fb48f8f05d8478bd5c5205c3221c8c7488efd51 100644 (file)
@@ -304,6 +304,8 @@ static void engine_activity_set_cpu_ts(struct xe_guc *guc, unsigned int index)
        struct engine_activity_group *eag = &engine_activity->eag[index];
        int i, j;
 
+       xe_gt_assert(guc_to_gt(guc), index < engine_activity->num_activity_group);
+
        for (i = 0; i < GUC_MAX_ENGINE_CLASSES; i++)
                for (j = 0; j < GUC_MAX_INSTANCES_PER_CLASS; j++)
                        eag->engine[i][j].last_cpu_ts = ktime_get();
@@ -374,8 +376,9 @@ static int engine_activity_enable_function_stats(struct xe_guc *guc, int num_vfs
                return ret;
        }
 
-       for (i = 0; i < engine_activity->num_functions; i++)
-               engine_activity_set_cpu_ts(guc, i + 1);
+       /* skip PF as it was already setup */
+       for (i = 1; i < engine_activity->num_functions; i++)
+               engine_activity_set_cpu_ts(guc, i);
 
        return 0;
 }