Skip to contentSkip to conversation

drm/xe/display: Fix memleak in display initialization

PATCH In review wangxiaoming321 5 participants 8 messages Jan 25, 2024 – Feb 1, 2024

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/xe/display: Fix memleak in display initialization

In the call stack xe_device_probe -> xe_display_init_nommio -> intel_power_domains_init Power_domains hasn't been cleaned up if return error,
which has do the clean in i915_driver_late_release call from i915_driver_probe.

unreferenced object 0xffff88811150ee00 (size 512):
comm "systemd-udevd", pid 506, jiffies 4294674198 (age 3605.560s) hex dump (first 32 bytes):

    10 b4 9d a0 ff ff ff ff ff ff ff ff ff ff ff ff  ................
    ff ff ff ff ff ff ff ff 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff8134b901>] __kmem_cache_alloc_node+0x1c1/0x2b0
    [<ffffffff812c98b2>] __kmalloc+0x52/0x150
    [<ffffffffa08b0033>] __set_power_wells+0xc3/0x360 [xe]
    [<ffffffffa08562fc>] xe_display_init_nommio+0x4c/0x70 [xe]
    [<ffffffffa07f0d1c>] xe_device_probe+0x3c/0x5a0 [xe]
    [<ffffffffa082e48f>] xe_pci_probe+0x33f/0x5a0 [xe]
    [<ffffffff817f2187>] local_pci_probe+0x47/0xa0
    [<ffffffff817f3db3>] pci_device_probe+0xc3/0x1f0
    [<ffffffff8192f2a2>] really_probe+0x1a2/0x410
    [<ffffffff8192f598>] __driver_probe_device+0x78/0x160
    [<ffffffff8192f6ae>] driver_probe_device+0x1e/0x90
    [<ffffffff8192f92a>] __driver_attach+0xda/0x1d0
    [<ffffffff8192c95c>] bus_for_each_dev+0x7c/0xd0
    [<ffffffff8192e159>] bus_add_driver+0x119/0x220
    [<ffffffff81930d00>] driver_register+0x60/0x120
    [<ffffffffa05e50a0>] 0xffffffffa05e50a0
---
 drivers/gpu/drm/xe/xe_display.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/gpu/drm/xe/xe_display.c +3 -1
index 74391d9b11ae..2725afba4afb 100644
@@ -146,8 +146,10 @@ int xe_display_init_nommio(struct xe_device *xe)
 	intel_detect_pch(xe);
 
 	err = intel_power_domains_init(xe);
-	if (err)
+	if (err) {
+		intel_power_domains_cleanup(xe);
 		return err;
+	}
 
 	return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
 }
Signature · 1 line
2.25.1
maintainer
#1 lore raw
Quoted 3 lines · On Thu, 25 Jan 2024, wangxiaoming321 <[email protected]> wrote:

In the call stack xe_device_probe -> xe_display_init_nommio -> intel_power_domains_init Power_domains hasn't been cleaned up if return error,
which has do the clean in i915_driver_late_release call from i915_driver_probe.

This has nothing to do with i915_*.

If intel_power_domains_init() returns an error, it should have cleaned up after itself, not force its caller to do that. If there's an issue, please fix it in intel_power_domains_init().

BR,
Jani.

Quoted 45 lines

unreferenced object 0xffff88811150ee00 (size 512):
comm "systemd-udevd", pid 506, jiffies 4294674198 (age 3605.560s) hex dump (first 32 bytes):

    10 b4 9d a0 ff ff ff ff ff ff ff ff ff ff ff ff  ................
    ff ff ff ff ff ff ff ff 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff8134b901>] __kmem_cache_alloc_node+0x1c1/0x2b0
    [<ffffffff812c98b2>] __kmalloc+0x52/0x150
    [<ffffffffa08b0033>] __set_power_wells+0xc3/0x360 [xe]
    [<ffffffffa08562fc>] xe_display_init_nommio+0x4c/0x70 [xe]
    [<ffffffffa07f0d1c>] xe_device_probe+0x3c/0x5a0 [xe]
    [<ffffffffa082e48f>] xe_pci_probe+0x33f/0x5a0 [xe]
    [<ffffffff817f2187>] local_pci_probe+0x47/0xa0
    [<ffffffff817f3db3>] pci_device_probe+0xc3/0x1f0
    [<ffffffff8192f2a2>] really_probe+0x1a2/0x410
    [<ffffffff8192f598>] __driver_probe_device+0x78/0x160
    [<ffffffff8192f6ae>] driver_probe_device+0x1e/0x90
    [<ffffffff8192f92a>] __driver_attach+0xda/0x1d0
    [<ffffffff8192c95c>] bus_for_each_dev+0x7c/0xd0
    [<ffffffff8192e159>] bus_add_driver+0x119/0x220
    [<ffffffff81930d00>] driver_register+0x60/0x120
    [<ffffffffa05e50a0>] 0xffffffffa05e50a0
---
 drivers/gpu/drm/xe/xe_display.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/gpu/drm/xe/xe_display.c +3 -1
index 74391d9b11ae..2725afba4afb 100644
@@ -146,8 +146,10 @@ int xe_display_init_nommio(struct xe_device *xe)
 	intel_detect_pch(xe);
 
 	err = intel_power_domains_init(xe);
-	if (err)
+	if (err) {
+		intel_power_domains_cleanup(xe);
 		return err;
+	}
 
 	return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
 }
Signature · 1 line
Jani Nikula, Intel
#2 lore raw

PATCH 1/1 drm/xe/display: Fix memleak in display initialization

In reply to wangxiaoming321

intel_power_domains_init has been called twice in xe_device_probe:
xe_device_probe -> xe_display_init_nommio -> intel_power_domains_init(xe) xe_device_probe -> xe_display_init_noirq -> intel_display_driver_probe_noirq -> intel_power_domains_init(i915)

It needs add a flag to avoid power_domains->power_wells double malloc.

unreferenced object 0xffff88811150ee00 (size 512):
comm "systemd-udevd", pid 506, jiffies 4294674198 (age 3605.560s) hex dump (first 32 bytes):

    10 b4 9d a0 ff ff ff ff ff ff ff ff ff ff ff ff  ................
    ff ff ff ff ff ff ff ff 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff8134b901>] __kmem_cache_alloc_node+0x1c1/0x2b0
    [<ffffffff812c98b2>] __kmalloc+0x52/0x150
    [<ffffffffa08b0033>] __set_power_wells+0xc3/0x360 [xe]
    [<ffffffffa08562fc>] xe_display_init_nommio+0x4c/0x70 [xe]
    [<ffffffffa07f0d1c>] xe_device_probe+0x3c/0x5a0 [xe]
    [<ffffffffa082e48f>] xe_pci_probe+0x33f/0x5a0 [xe]
    [<ffffffff817f2187>] local_pci_probe+0x47/0xa0
    [<ffffffff817f3db3>] pci_device_probe+0xc3/0x1f0
    [<ffffffff8192f2a2>] really_probe+0x1a2/0x410
    [<ffffffff8192f598>] __driver_probe_device+0x78/0x160
    [<ffffffff8192f6ae>] driver_probe_device+0x1e/0x90
    [<ffffffff8192f92a>] __driver_attach+0xda/0x1d0
    [<ffffffff8192c95c>] bus_for_each_dev+0x7c/0xd0
    [<ffffffff8192e159>] bus_add_driver+0x119/0x220
    [<ffffffff81930d00>] driver_register+0x60/0x120
    [<ffffffffa05e50a0>] 0xffffffffa05e50a0
---
 drivers/gpu/drm/i915/display/intel_display_power.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/gpu/drm/i915/display/intel_display_power.c +8 -0
index bf9685acf75a..3b48a1cb7c54 100644
@@ -36,6 +36,8 @@
 	for_each_power_well_reverse(__dev_priv, __power_well)		        \
 		for_each_if(test_bit((__domain), (__power_well)->domains.bits))
 
+static int intel_power_domains_init_flag = 0;
+
 const char *
 intel_display_power_domain_str(enum intel_display_power_domain domain)
 {
@@ -1016,6 +1018,11 @@ int intel_power_domains_init(struct drm_i915_private *dev_priv)
 {
 	struct i915_power_domains *power_domains = &dev_priv->display.power.domains;
 
+	if(intel_power_domains_init_flag == 1)
+		return 0;
+
+	intel_power_domains_init_flag++;
+
 	dev_priv->display.params.disable_power_well =
 		sanitize_disable_power_well_option(dev_priv,
 						   dev_priv->display.params.disable_power_well);
@@ -1041,6 +1048,7 @@ int intel_power_domains_init(struct drm_i915_private *dev_priv)
  */
 void intel_power_domains_cleanup(struct drm_i915_private *dev_priv)
 {
+	intel_power_domains_init_flag = 0;
 	intel_display_power_map_cleanup(&dev_priv->display.power.domains);
 }
 
Signature · 1 line
2.25.1
#3 lore raw
Quoted 4 lines · On Fri, Jan 26, 2024 at 10:34:33PM +0800, wangxiaoming321 wrote:

intel_power_domains_init has been called twice in xe_device_probe:
xe_device_probe -> xe_display_init_nommio -> intel_power_domains_init(xe) xe_device_probe -> xe_display_init_noirq -> intel_display_driver_probe_noirq -> intel_power_domains_init(i915)

Why are you calling it twice?

Quoted 45 lines

It needs add a flag to avoid power_domains->power_wells double malloc.

unreferenced object 0xffff88811150ee00 (size 512):
comm "systemd-udevd", pid 506, jiffies 4294674198 (age 3605.560s) hex dump (first 32 bytes):

    10 b4 9d a0 ff ff ff ff ff ff ff ff ff ff ff ff  ................
    ff ff ff ff ff ff ff ff 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff8134b901>] __kmem_cache_alloc_node+0x1c1/0x2b0
    [<ffffffff812c98b2>] __kmalloc+0x52/0x150
    [<ffffffffa08b0033>] __set_power_wells+0xc3/0x360 [xe]
    [<ffffffffa08562fc>] xe_display_init_nommio+0x4c/0x70 [xe]
    [<ffffffffa07f0d1c>] xe_device_probe+0x3c/0x5a0 [xe]
    [<ffffffffa082e48f>] xe_pci_probe+0x33f/0x5a0 [xe]
    [<ffffffff817f2187>] local_pci_probe+0x47/0xa0
    [<ffffffff817f3db3>] pci_device_probe+0xc3/0x1f0
    [<ffffffff8192f2a2>] really_probe+0x1a2/0x410
    [<ffffffff8192f598>] __driver_probe_device+0x78/0x160
    [<ffffffff8192f6ae>] driver_probe_device+0x1e/0x90
    [<ffffffff8192f92a>] __driver_attach+0xda/0x1d0
    [<ffffffff8192c95c>] bus_for_each_dev+0x7c/0xd0
    [<ffffffff8192e159>] bus_add_driver+0x119/0x220
    [<ffffffff81930d00>] driver_register+0x60/0x120
    [<ffffffffa05e50a0>] 0xffffffffa05e50a0
---
 drivers/gpu/drm/i915/display/intel_display_power.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/gpu/drm/i915/display/intel_display_power.c +2 -0
index bf9685acf75a..3b48a1cb7c54 100644
@@ -36,6 +36,8 @@
 	for_each_power_well_reverse(__dev_priv, __power_well)		        \
 		for_each_if(test_bit((__domain), (__power_well)->domains.bits))
 
+static int intel_power_domains_init_flag = 0;
+
 const char *
 intel_display_power_domain_str(enum intel_display_power_domain domain)
 {

drivers/gpu/drm/i915/display/intel_display_power.c int intel_power_domains_init(struct drm_i915_private *dev_priv) · line 1024

Quoted 7 lines
drivers/gpu/drm/i915/display/intel_display_power.c +4 -0
@@ -1016,6 +1018,11 @@ int intel_power_domains_init(struct drm_i915_private *dev_priv)
 {
 	struct i915_power_domains *power_domains = &dev_priv->display.power.domains;
 
+	if(intel_power_domains_init_flag == 1)
+		return 0;
+
+	intel_power_domains_init_flag++;

Consider what happens when you have multiple Intel GPUs in the system...

Quoted 14 lines
drivers/gpu/drm/i915/display/intel_display_power.c +2 -0
+
 	dev_priv->display.params.disable_power_well =
 		sanitize_disable_power_well_option(dev_priv,
 						   dev_priv->display.params.disable_power_well);
@@ -1041,6 +1048,7 @@ int intel_power_domains_init(struct drm_i915_private *dev_priv)
  */
 void intel_power_domains_cleanup(struct drm_i915_private *dev_priv)
 {
+	intel_power_domains_init_flag = 0;
 	intel_display_power_map_cleanup(&dev_priv->display.power.domains);
 }
 
Signature · 1 line
2.25.1
Signature · 2 lines
Ville Syrjälä
Intel
#4 lore raw

PATCH 1/1 drm/xe/display: Fix memleak in display initialization

In reply to wangxiaoming321

intel_power_domains_init has been called twice in xe_device_probe:
xe_device_probe -> xe_display_init_nommio -> intel_power_domains_init(xe) xe_device_probe -> xe_display_init_noirq -> intel_display_driver_probe_noirq -> intel_power_domains_init(i915)

It needs remove one to avoid power_domains->power_wells double malloc.

unreferenced object 0xffff88811150ee00 (size 512):
comm "systemd-udevd", pid 506, jiffies 4294674198 (age 3605.560s) hex dump (first 32 bytes):

    10 b4 9d a0 ff ff ff ff ff ff ff ff ff ff ff ff  ................
    ff ff ff ff ff ff ff ff 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff8134b901>] __kmem_cache_alloc_node+0x1c1/0x2b0
    [<ffffffff812c98b2>] __kmalloc+0x52/0x150
    [<ffffffffa08b0033>] __set_power_wells+0xc3/0x360 [xe]
    [<ffffffffa08562fc>] xe_display_init_nommio+0x4c/0x70 [xe]
    [<ffffffffa07f0d1c>] xe_device_probe+0x3c/0x5a0 [xe]
    [<ffffffffa082e48f>] xe_pci_probe+0x33f/0x5a0 [xe]
    [<ffffffff817f2187>] local_pci_probe+0x47/0xa0
    [<ffffffff817f3db3>] pci_device_probe+0xc3/0x1f0
    [<ffffffff8192f2a2>] really_probe+0x1a2/0x410
    [<ffffffff8192f598>] __driver_probe_device+0x78/0x160
    [<ffffffff8192f6ae>] driver_probe_device+0x1e/0x90
    [<ffffffff8192f92a>] __driver_attach+0xda/0x1d0
    [<ffffffff8192c95c>] bus_for_each_dev+0x7c/0xd0
    [<ffffffff8192e159>] bus_add_driver+0x119/0x220
    [<ffffffff81930d00>] driver_register+0x60/0x120
    [<ffffffffa05e50a0>] 0xffffffffa05e50a0
---
 drivers/gpu/drm/xe/xe_display.c | 6 ------
 1 file changed, 6 deletions(-)
drivers/gpu/drm/xe/xe_display.c +0 -6
index 74391d9b11ae..e4db069f0db3 100644
@@ -134,8 +134,6 @@ static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
 
 int xe_display_init_nommio(struct xe_device *xe)
 {
-	int err;
-
 	if (!xe->info.enable_display)
 		return 0;
 
@@ -145,10 +143,6 @@ int xe_display_init_nommio(struct xe_device *xe)
 	/* This must be called before any calls to HAS_PCH_* */
 	intel_detect_pch(xe);
 
-	err = intel_power_domains_init(xe);
-	if (err)
-		return err;
-
 	return drmm_add_action_or_reset(&xe->drm, xe_display_fini_nommio, xe);
 }
 
Signature · 1 line
2.25.1
#5 lore raw

+Jani

Quoted 4 lines · On Fri, Jan 26, 2024 at 11:34:53PM +0800, wangxiaoming321 wrote:

intel_power_domains_init has been called twice in xe_device_probe:
xe_device_probe -> xe_display_init_nommio -> intel_power_domains_init(xe) xe_device_probe -> xe_display_init_noirq -> intel_display_driver_probe_noirq -> intel_power_domains_init(i915)

ok, once upon a time intel_power_domains_init() was called by the driver initialization code and not initialized inside the display. I think. Now it's part of the display probe and we never updated the xe side.

Quoted 9 lines

It needs remove one to avoid power_domains->power_wells double malloc.

unreferenced object 0xffff88811150ee00 (size 512):
comm "systemd-udevd", pid 506, jiffies 4294674198 (age 3605.560s) hex dump (first 32 bytes):
10 b4 9d a0 ff ff ff ff ff ff ff ff ff ff ff ff ................ ff ff ff ff ff ff ff ff 00 00 00 00 00 00 00 00 ................
backtrace:

Quoted 17 lines
   [<ffffffff8134b901>] __kmem_cache_alloc_node+0x1c1/0x2b0
   [<ffffffff812c98b2>] __kmalloc+0x52/0x150
   [<ffffffffa08b0033>] __set_power_wells+0xc3/0x360 [xe]
   [<ffffffffa08562fc>] xe_display_init_nommio+0x4c/0x70 [xe]
   [<ffffffffa07f0d1c>] xe_device_probe+0x3c/0x5a0 [xe]
   [<ffffffffa082e48f>] xe_pci_probe+0x33f/0x5a0 [xe]
   [<ffffffff817f2187>] local_pci_probe+0x47/0xa0
   [<ffffffff817f3db3>] pci_device_probe+0xc3/0x1f0
   [<ffffffff8192f2a2>] really_probe+0x1a2/0x410
   [<ffffffff8192f598>] __driver_probe_device+0x78/0x160
   [<ffffffff8192f6ae>] driver_probe_device+0x1e/0x90
   [<ffffffff8192f92a>] __driver_attach+0xda/0x1d0
   [<ffffffff8192c95c>] bus_for_each_dev+0x7c/0xd0
   [<ffffffff8192e159>] bus_add_driver+0x119/0x220
   [<ffffffff81930d00>] driver_register+0x60/0x120
   [<ffffffffa05e50a0>] 0xffffffffa05e50a0

This will need a Fixes trailer. This seems to be a suitable one:

Fixes 44e694958b95 ("drm/xe/display: Implement display support")
Quoted 19 lines

---
drivers/gpu/drm/xe/xe_display.c | 6 ------

1 file changed, 6 deletions(-)
drivers/gpu/drm/xe/xe_display.c +0 -2
index 74391d9b11ae..e4db069f0db3 100644
@@ -134,8 +134,6 @@ static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
 
 int xe_display_init_nommio(struct xe_device *xe)
 {
-	int err;
-
 	if (!xe->info.enable_display)
 		return 0;
 

drivers/gpu/drm/xe/xe_display.c int xe_display_init_nommio(struct xe_device *xe) · line 150

Quoted 6 lines
drivers/gpu/drm/xe/xe_display.c +0 -3
@@ -145,10 +143,6 @@ int xe_display_init_nommio(struct xe_device *xe)
 	/* This must be called before any calls to HAS_PCH_* */
 	intel_detect_pch(xe);
 
-	err = intel_power_domains_init(xe);
-	if (err)
-		return err;

xe_display_init_nommio() has xe_display_fini_nommio() as its destructor counter part. Unfortunately display side looks wrong as it does:

init:

	intel_display_driver_probe_noirq() -> intel_power_domains_init()

destroy:

	i915_driver_late_release() -> intel_power_domains_cleanup()

I think leaving intel_power_domains_cleanup() as is for now so it's called by xe works, but this needs to go through CI, which apparently this series didn't go. I re-triggered it.

+Jani if he thinks this can be changed in another way or already have the complete solution.

Lucas De Marchi

maintainer
#6 lore raw
Quoted 76 lines · Lucas De Marchi

+Jani

Quoted 4 lines · On Fri, Jan 26, 2024 at 11:34:53PM +0800, wangxiaoming321 wrote:

intel_power_domains_init has been called twice in xe_device_probe:
xe_device_probe -> xe_display_init_nommio -> intel_power_domains_init(xe) xe_device_probe -> xe_display_init_noirq -> intel_display_driver_probe_noirq -> intel_power_domains_init(i915)

ok, once upon a time intel_power_domains_init() was called by the driver initialization code and not initialized inside the display. I think. Now it's part of the display probe and we never updated the xe side.

Quoted 9 lines

It needs remove one to avoid power_domains->power_wells double malloc.

unreferenced object 0xffff88811150ee00 (size 512):
comm "systemd-udevd", pid 506, jiffies 4294674198 (age 3605.560s) hex dump (first 32 bytes):
10 b4 9d a0 ff ff ff ff ff ff ff ff ff ff ff ff ................ ff ff ff ff ff ff ff ff 00 00 00 00 00 00 00 00 ................
backtrace:

Quoted 17 lines
   [<ffffffff8134b901>] __kmem_cache_alloc_node+0x1c1/0x2b0
   [<ffffffff812c98b2>] __kmalloc+0x52/0x150
   [<ffffffffa08b0033>] __set_power_wells+0xc3/0x360 [xe]
   [<ffffffffa08562fc>] xe_display_init_nommio+0x4c/0x70 [xe]
   [<ffffffffa07f0d1c>] xe_device_probe+0x3c/0x5a0 [xe]
   [<ffffffffa082e48f>] xe_pci_probe+0x33f/0x5a0 [xe]
   [<ffffffff817f2187>] local_pci_probe+0x47/0xa0
   [<ffffffff817f3db3>] pci_device_probe+0xc3/0x1f0
   [<ffffffff8192f2a2>] really_probe+0x1a2/0x410
   [<ffffffff8192f598>] __driver_probe_device+0x78/0x160
   [<ffffffff8192f6ae>] driver_probe_device+0x1e/0x90
   [<ffffffff8192f92a>] __driver_attach+0xda/0x1d0
   [<ffffffff8192c95c>] bus_for_each_dev+0x7c/0xd0
   [<ffffffff8192e159>] bus_add_driver+0x119/0x220
   [<ffffffff81930d00>] driver_register+0x60/0x120
   [<ffffffffa05e50a0>] 0xffffffffa05e50a0

This will need a Fixes trailer. This seems to be a suitable one:

Fixes 44e694958b95 ("drm/xe/display: Implement display support")
Quoted 19 lines

---
drivers/gpu/drm/xe/xe_display.c | 6 ------

1 file changed, 6 deletions(-)
drivers/gpu/drm/xe/xe_display.c +0 -2
index 74391d9b11ae..e4db069f0db3 100644
@@ -134,8 +134,6 @@ static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
 
 int xe_display_init_nommio(struct xe_device *xe)
 {
-	int err;
-
 	if (!xe->info.enable_display)
 		return 0;
 

drivers/gpu/drm/xe/xe_display.c int xe_display_init_nommio(struct xe_device *xe) · line 150

Quoted 6 lines
drivers/gpu/drm/xe/xe_display.c +0 -3
@@ -145,10 +143,6 @@ int xe_display_init_nommio(struct xe_device *xe)
 	/* This must be called before any calls to HAS_PCH_* */
 	intel_detect_pch(xe);
 
-	err = intel_power_domains_init(xe);
-	if (err)
-		return err;

xe_display_init_nommio() has xe_display_fini_nommio() as its destructor counter part. Unfortunately display side looks wrong as it does:

init:

	intel_display_driver_probe_noirq() -> intel_power_domains_init()

destroy:

Quoted 8 lines
	i915_driver_late_release() -> intel_power_domains_cleanup()

I think leaving intel_power_domains_cleanup() as is for now so it's called by xe works, but this needs to go through CI, which apparently this series didn't go. I re-triggered it.

+Jani if he thinks this can be changed in another way or already have the complete solution.

I don't. But it is and will be a recurring problem. i915 and xe core drivers should handle display init and cleanup the same way. But currently i915 goes on to call e.g. intel_power_domains_cleanup() directly from top level driver code. There are other examples.

And we seem to have recently added *more*. See e.g. bd738d859e71 ("drm/i915: Prevent modesets during driver init/shutdown").

BR,
Jani.

Signature · 1 line
Jani Nikula, Intel
#7 lore raw
Quoted 86 lines · Jani Nikula
Quoted 76 lines · Lucas De Marchi

+Jani

Quoted 4 lines · On Fri, Jan 26, 2024 at 11:34:53PM +0800, wangxiaoming321 wrote:

intel_power_domains_init has been called twice in xe_device_probe:
xe_device_probe -> xe_display_init_nommio -> intel_power_domains_init(xe) xe_device_probe -> xe_display_init_noirq -> intel_display_driver_probe_noirq -> intel_power_domains_init(i915)

ok, once upon a time intel_power_domains_init() was called by the driver initialization code and not initialized inside the display. I think. Now it's part of the display probe and we never updated the xe side.

Quoted 17 lines

It needs remove one to avoid power_domains->power_wells double malloc.

unreferenced object 0xffff88811150ee00 (size 512):
comm "systemd-udevd", pid 506, jiffies 4294674198 (age 3605.560s) hex dump (first 32 bytes):

    10 b4 9d a0 ff ff ff ff ff ff ff ff ff ff ff ff  ................
    ff ff ff ff ff ff ff ff 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff8134b901>] __kmem_cache_alloc_node+0x1c1/0x2b0
    [<ffffffff812c98b2>] __kmalloc+0x52/0x150
    [<ffffffffa08b0033>] __set_power_wells+0xc3/0x360 [xe]
    [<ffffffffa08562fc>] xe_display_init_nommio+0x4c/0x70 [xe]
    [<ffffffffa07f0d1c>] xe_device_probe+0x3c/0x5a0 [xe]
    [<ffffffffa082e48f>] xe_pci_probe+0x33f/0x5a0 [xe]
    [<ffffffff817f2187>] local_pci_probe+0x47/0xa0
    [<ffffffff817f3db3>] pci_device_probe+0xc3/0x1f0
Quoted 9 lines
    [<ffffffff8192f2a2>] really_probe+0x1a2/0x410
    [<ffffffff8192f598>] __driver_probe_device+0x78/0x160
    [<ffffffff8192f6ae>] driver_probe_device+0x1e/0x90
    [<ffffffff8192f92a>] __driver_attach+0xda/0x1d0
    [<ffffffff8192c95c>] bus_for_each_dev+0x7c/0xd0
    [<ffffffff8192e159>] bus_add_driver+0x119/0x220
    [<ffffffff81930d00>] driver_register+0x60/0x120
    [<ffffffffa05e50a0>] 0xffffffffa05e50a0

This will need a Fixes trailer. This seems to be a suitable one:

Fixes 44e694958b95 ("drm/xe/display: Implement display support")
Quoted 19 lines

---
drivers/gpu/drm/xe/xe_display.c | 6 ------

1 file changed, 6 deletions(-)
drivers/gpu/drm/xe/xe_display.c +0 -2
index 74391d9b11ae..e4db069f0db3 100644
@@ -134,8 +134,6 @@ static void xe_display_fini_nommio(struct drm_device *dev, void *dummy)
 
 int xe_display_init_nommio(struct xe_device *xe)
 {
-	int err;
-
 	if (!xe->info.enable_display)
 		return 0;
 

drivers/gpu/drm/xe/xe_display.c int xe_display_init_nommio(struct xe_device *xe) · line 150

Quoted 6 lines
drivers/gpu/drm/xe/xe_display.c +0 -3
@@ -145,10 +143,6 @@ int xe_display_init_nommio(struct xe_device *xe)
 	/* This must be called before any calls to HAS_PCH_* */
 	intel_detect_pch(xe);
 
-	err = intel_power_domains_init(xe);
-	if (err)
-		return err;

xe_display_init_nommio() has xe_display_fini_nommio() as its destructor counter part. Unfortunately display side looks wrong as it does:

init:

	intel_display_driver_probe_noirq() -> intel_power_domains_init()

destroy:

Quoted 8 lines
	i915_driver_late_release() -> intel_power_domains_cleanup()

I think leaving intel_power_domains_cleanup() as is for now so it's called by xe works, but this needs to go through CI, which apparently this series didn't go. I re-triggered it.

+Jani if he thinks this can be changed in another way or already have the complete solution.

Quoted 7 lines

I don't. But it is and will be a recurring problem. i915 and xe core drivers should handle display init and cleanup the same way. But currently i915 goes on to call e.g. intel_power_domains_cleanup() directly from top level driver code. There are other examples.

And we seem to have recently added *more*. See e.g. bd738d859e71 ("drm/i915: Prevent modesets during driver init/shutdown").

That commit seems terrible Should we instead not only enable any code that can cause modesets after it's safe to do so?

Cheers,
~Maarten