Sergey Senozhatsky
2019-May-17 06:13 UTC
[Nouveau] drm/nouveau/core/memory: kmemleak 684 new suspected memory leaks
Hello,
5.1.0-next-20190517
I'm looking at quite a lot of kmemleak reports coming from
drm/nouveau/core/memory, all of which are:
unreferenced object 0xffff8deec27c4ac0 (size 16):
comm "Web Content", pid 5309, jiffies 4309675011 (age 68.076s)
hex dump (first 16 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<0000000081f2894f>] nvkm_memory_tags_get+0x8e/0x130
[<000000007cd7c0bc>] gf100_vmm_valid+0x196/0x2f0
[<0000000070cc6d67>] nvkm_vmm_map+0xa8/0x360
[<00000000ab678644>] nvkm_vram_map+0x48/0x50
[<00000000d8176378>] nvkm_uvmm_mthd+0x658/0x770
[<00000000463fca5a>] nvkm_ioctl+0xdf/0x177
[<000000000afc4996>] nvif_object_mthd+0xd4/0x100
[<000000002f7a7385>] nvif_vmm_map+0xeb/0x100
[<00000000ef2537ed>] nouveau_mem_map+0x79/0xd0
[<0000000014ddc0cf>] nouveau_vma_new+0x19d/0x1c0
[<00000000f99888a1>] nouveau_gem_object_open+0xd4/0x140
[<000000009cd25861>] drm_gem_handle_create_tail+0xe3/0x160
[<00000000191784d9>] nouveau_gem_ioctl_new+0x6e/0xd0
[<00000000159678df>] drm_ioctl_kernel+0x8c/0xd0
[<00000000fbaa6154>] drm_ioctl+0x1c4/0x360
[<000000006833fe15>] nouveau_drm_ioctl+0x63/0xb0
Wondering if those are real leaks or just false positives.
For now I marked `tags' as kmemleak_not_leak(); but most
likely it's utterly wrong.
Any thoughts?
---
drivers/gpu/drm/nouveau/nvkm/core/memory.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/gpu/drm/nouveau/nvkm/core/memory.c
b/drivers/gpu/drm/nouveau/nvkm/core/memory.c
index e85a08ecd9da..cd46f54c5c32 100644
--- a/drivers/gpu/drm/nouveau/nvkm/core/memory.c
+++ b/drivers/gpu/drm/nouveau/nvkm/core/memory.c
@@ -25,6 +25,7 @@
#include <core/mm.h>
#include <subdev/fb.h>
#include <subdev/instmem.h>
+#include <linux/kmemleak.h>
void
nvkm_memory_tags_put(struct nvkm_memory *memory, struct nvkm_device *device,
@@ -92,6 +93,7 @@ nvkm_memory_tags_get(struct nvkm_memory *memory, struct
nvkm_device *device,
refcount_set(&tags->refcount, 1);
mutex_unlock(&fb->subdev.mutex);
+ kmemleak_not_leak(tags);
*ptags = tags;
return 0;
}
--
2.21.0
Sergey Senozhatsky
2019-May-17 06:31 UTC
[Nouveau] drm/nouveau/core/memory: kmemleak 684 new suspected memory leaks
On (05/17/19 15:13), Sergey Senozhatsky wrote:> 5.1.0-next-20190517 > > I'm looking at quite a lot of kmemleak reports coming from > drm/nouveau/core/memory, all of which are: > > unreferenced object 0xffff8deec27c4ac0 (size 16): > comm "Web Content", pid 5309, jiffies 4309675011 (age 68.076s) > hex dump (first 16 bytes): > 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ > backtrace: > [<0000000081f2894f>] nvkm_memory_tags_get+0x8e/0x130 > [<000000007cd7c0bc>] gf100_vmm_valid+0x196/0x2f0 > [<0000000070cc6d67>] nvkm_vmm_map+0xa8/0x360 > [<00000000ab678644>] nvkm_vram_map+0x48/0x50 > [<00000000d8176378>] nvkm_uvmm_mthd+0x658/0x770 > [<00000000463fca5a>] nvkm_ioctl+0xdf/0x177 > [<000000000afc4996>] nvif_object_mthd+0xd4/0x100 > [<000000002f7a7385>] nvif_vmm_map+0xeb/0x100 > [<00000000ef2537ed>] nouveau_mem_map+0x79/0xd0 > [<0000000014ddc0cf>] nouveau_vma_new+0x19d/0x1c0 > [<00000000f99888a1>] nouveau_gem_object_open+0xd4/0x140 > [<000000009cd25861>] drm_gem_handle_create_tail+0xe3/0x160 > [<00000000191784d9>] nouveau_gem_ioctl_new+0x6e/0xd0 > [<00000000159678df>] drm_ioctl_kernel+0x8c/0xd0 > [<00000000fbaa6154>] drm_ioctl+0x1c4/0x360 > [<000000006833fe15>] nouveau_drm_ioctl+0x63/0xb0Yet another one (4 leaks), but this looks more like a real leak. unreferenced object 0xffff8f1e0cbbe840 (size 192): comm "swapper/0", pid 1, jiffies 4294668445 (age 742.639s) hex dump (first 32 bytes): 00 90 89 0c 1e 8f ff ff 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ backtrace: [<000000006933ed2b>] nouveau_conn_reset+0x20/0xb0 [<00000000572e2e30>] nouveau_connector_create+0x356/0x54c [<000000008a6a13cd>] nv50_display_create+0x2fb/0x917 [<000000007fab0a58>] nouveau_display_create+0x3e6/0x600 [<000000008b8644c8>] nouveau_drm_device_init+0x149/0x6b0 [<000000004fd78a1f>] nouveau_drm_probe+0x263/0x2b0 [<00000000357716ef>] pci_device_probe+0xa3/0x110 [<00000000061d40e4>] really_probe+0xd3/0x240 [<000000000ade44b6>] driver_probe_device+0x50/0xc0 [<000000009cd0024c>] device_driver_attach+0x53/0x60 [<00000000b11ab0bb>] __driver_attach+0x4c/0xb0 [<0000000016d8457f>] bus_for_each_dev+0x66/0x90 [<00000000f2855f5e>] bus_add_driver+0x171/0x1c0 [<0000000021c08fc1>] driver_register+0x6c/0xaf [<0000000086357843>] do_one_initcall+0x36/0x1d4 [<00000000a6be055a>] kernel_init_freeable+0x1bf/0x24f Seems that connector ->state is not fully destroyed. --- drivers/gpu/drm/nouveau/nouveau_connector.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.c b/drivers/gpu/drm/nouveau/nouveau_connector.c index 4116ee62adaf..caec1737a7de 100644 --- a/drivers/gpu/drm/nouveau/nouveau_connector.c +++ b/drivers/gpu/drm/nouveau/nouveau_connector.c @@ -251,8 +251,10 @@ nouveau_conn_reset(struct drm_connector *connector) if (WARN_ON(!(asyc = kzalloc(sizeof(*asyc), GFP_KERNEL)))) return; - if (connector->state) + if (connector->state) { __drm_atomic_helper_connector_destroy_state(connector->state); + kfree(connector->state); + } __drm_atomic_helper_connector_reset(connector, &asyc->state); asyc->dither.mode = DITHERING_MODE_AUTO; asyc->dither.depth = DITHERING_DEPTH_AUTO; -- 2.21.0
Sergey Senozhatsky
2019-May-17 06:45 UTC
[Nouveau] drm/nouveau/core/memory: kmemleak 684 new suspected memory leaks
On (05/17/19 15:31), Sergey Senozhatsky wrote:> > backtrace: > > [<0000000081f2894f>] nvkm_memory_tags_get+0x8e/0x130 > > [<000000007cd7c0bc>] gf100_vmm_valid+0x196/0x2f0 > > [<0000000070cc6d67>] nvkm_vmm_map+0xa8/0x360 > > [<00000000ab678644>] nvkm_vram_map+0x48/0x50 > > [<00000000d8176378>] nvkm_uvmm_mthd+0x658/0x770 > > [<00000000463fca5a>] nvkm_ioctl+0xdf/0x177 > > [<000000000afc4996>] nvif_object_mthd+0xd4/0x100 > > [<000000002f7a7385>] nvif_vmm_map+0xeb/0x100 > > [<00000000ef2537ed>] nouveau_mem_map+0x79/0xd0 > > [<0000000014ddc0cf>] nouveau_vma_new+0x19d/0x1c0 > > [<00000000f99888a1>] nouveau_gem_object_open+0xd4/0x140 > > [<000000009cd25861>] drm_gem_handle_create_tail+0xe3/0x160 > > [<00000000191784d9>] nouveau_gem_ioctl_new+0x6e/0xd0 > > [<00000000159678df>] drm_ioctl_kernel+0x8c/0xd0 > > [<00000000fbaa6154>] drm_ioctl+0x1c4/0x360 > > [<000000006833fe15>] nouveau_drm_ioctl+0x63/0xb0[..]> backtrace: > [<000000006933ed2b>] nouveau_conn_reset+0x20/0xb0 > [<00000000572e2e30>] nouveau_connector_create+0x356/0x54c > [<000000008a6a13cd>] nv50_display_create+0x2fb/0x917 > [<000000007fab0a58>] nouveau_display_create+0x3e6/0x600 > [<000000008b8644c8>] nouveau_drm_device_init+0x149/0x6b0 > [<000000004fd78a1f>] nouveau_drm_probe+0x263/0x2b0 > [<00000000357716ef>] pci_device_probe+0xa3/0x110 > [<00000000061d40e4>] really_probe+0xd3/0x240 > [<000000000ade44b6>] driver_probe_device+0x50/0xc0 > [<000000009cd0024c>] device_driver_attach+0x53/0x60 > [<00000000b11ab0bb>] __driver_attach+0x4c/0xb0 > [<0000000016d8457f>] bus_for_each_dev+0x66/0x90 > [<00000000f2855f5e>] bus_add_driver+0x171/0x1c0 > [<0000000021c08fc1>] driver_register+0x6c/0xaf > [<0000000086357843>] do_one_initcall+0x36/0x1d4 > [<00000000a6be055a>] kernel_init_freeable+0x1bf/0x24fAnd one more: iccsense->rail = kmalloc_array(cnt, sizeof(struct pwr_rail_t), GFP_KERNEL); unreferenced object 0xffff94e5ccdc7600 (size 96): comm "swapper/0", pid 1, jiffies 4294667774 (age 913.205s) hex dump (first 32 bytes): 00 00 00 cc e5 94 ff ff 00 00 00 00 00 00 00 00 ................ 04 00 f1 ff 00 00 00 00 00 00 00 00 00 00 00 00 ................ backtrace: [<00000000242abcb3>] nvbios_iccsense_parse+0xdc/0x250 [<00000000b5c70490>] nvkm_iccsense_oneinit+0x55/0x370 [<0000000020e0a743>] nvkm_subdev_init+0x53/0xd0 [<000000004d8c6ef1>] nvkm_device_init+0x10d/0x190 [<00000000bd7a4da4>] nvkm_udevice_init+0x41/0x60 [<0000000047effcfc>] nvkm_object_init+0x3e/0x100 [<000000006d6bad21>] nvkm_ioctl_new+0x145/0x1e0 [<00000000fc4e7e48>] nvkm_ioctl+0xdf/0x177 [<000000004cdc9cf8>] nvif_object_init+0xd6/0x130 [<000000001637584b>] nvif_device_init+0xe/0x50 [<00000000830683d4>] nouveau_cli_init+0x17d/0x410 [<00000000fd93c306>] nouveau_drm_device_init+0x55/0x6b0 [<000000007bc74e3f>] nouveau_drm_probe+0x263/0x2b0 [<000000000f94f913>] pci_device_probe+0xa3/0x110 [<00000000f82d21be>] really_probe+0xd3/0x240 [<00000000a8b8b02e>] driver_probe_device+0x50/0xc0 -ss
Sergey Senozhatsky
2019-May-17 07:27 UTC
[Nouveau] drm/nouveau/core/memory: kmemleak 684 new suspected memory leaks
On (05/17/19 15:13), Sergey Senozhatsky wrote:> ... but most likely it's utterly wrong. >JFI, I removed kmemleak annotation and added the following thing: @@ -360,6 +360,7 @@ gp100_vmm_valid(struct nvkm_vmm *vmm, void *argv, u32 argc, return -EINVAL; } + kfree(map->tags); ret = nvkm_memory_tags_get(memory, device, tags, nvkm_ltc_tags_clear, &map->tags); -ss
Sergey Senozhatsky
2019-May-17 07:35 UTC
[Nouveau] drm/nouveau/core/memory: kmemleak 684 new suspected memory leaks
On (05/17/19 16:27), Sergey Senozhatsky wrote:> > ... but most likely it's utterly wrong. > > > > JFI, I removed kmemleak annotationmeeehhh.... kmemleak: 2046 new suspected memory leaks (see /sys/kernel/debug/kmemleak) unreferenced object 0xffff95cbea4e6060 (size 16): comm "Web Content", pid 1191, jiffies 4294795669 (age 735.950s) hex dump (first 16 bytes): 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ backtrace: [<00000000d0781ea4>] nvkm_memory_tags_get+0x8e/0x130 [<00000000061f3c89>] gf100_vmm_valid+0x196/0x2f0 [<00000000d79084b7>] nvkm_vmm_map+0xa8/0x360 [<00000000e3174e33>] nvkm_vram_map+0x48/0x50 [<00000000006adddb>] nvkm_uvmm_mthd+0x658/0x770 [<00000000b36f3a8b>] nvkm_ioctl+0xdf/0x177 [<0000000003acea2c>] nvif_object_mthd+0xd4/0x100 [<0000000033824292>] nvif_vmm_map+0xeb/0x100 [<00000000537f8629>] nouveau_mem_map+0x79/0xd0 [<00000000c3b20b73>] nouveau_vma_new+0x19d/0x1c0 [<00000000dc91383f>] nouveau_gem_object_open+0xd4/0x140 [<000000005a53123b>] drm_gem_handle_create_tail+0xe3/0x160 [<00000000e733f5a8>] nouveau_gem_ioctl_new+0x6e/0xd0 [<00000000b5bebef5>] drm_ioctl_kernel+0x8c/0xd0 [<000000004f28d8a6>] drm_ioctl+0x1c4/0x360 [<00000000b08b5723>] nouveau_drm_ioctl+0x63/0xb0 -ss
Reasonably Related Threads
- drm/nouveau/core/memory: kmemleak 684 new suspected memory leaks
- drm/nouveau/core/memory: kmemleak 684 new suspected memory leaks
- [PATCH 2/2] drm/nouveau: Fix drm-core using atomic code-paths on pre-nv50 hardware
- [PATCH v2 2/2] drm/nouveau: Fix drm-core using atomic code-paths on pre-nv50 hardware
- [PATCH AUTOSEL 5.4 37/52] drm/nouveau: Fix drm-core using atomic code-paths on pre-nv50 hardware