On 3/6/24 20:49, Pierre-Louis Dourneau wrote: > > From: Arnaud Lahache <a.lahache@xxxxxxxxxx> > > Fixes a memory leak occurring on each modeset update. > > Running a program such as libdrm's modetest[0] with this driver exhausts > all available memory after a few minutes. Enabling kmemleak yields a series > of such leak reports: > > unreferenced object 0xc21acf40 (size 64): > comm "modetest", pid 210, jiffies 4294942460 (age 331.240s) > hex dump (first 32 bytes): > 00 a0 a1 c1 01 00 00 00 ff ff ff ff 4c cf 1a c2 ............L... > 4c cf 1a c2 ff ff ff ff 58 cf 1a c2 58 cf 1a c2 L.......X...X... > backtrace: > [<d68b3e09>] kmalloc_trace+0x18/0x24 > [<f858a020>] drm_atomic_helper_setup_commit+0x1e0/0x7e0 > [<26e8ab04>] drm_atomic_helper_commit+0x40/0x160 > [<49708b0c>] drm_atomic_commit+0xa8/0xf0 > [<e58c2942>] drm_mode_obj_set_property_ioctl+0x154/0x3d8 > [<5e97e57d>] drm_ioctl+0x200/0x3c4 > [<ed514ba1>] sys_ioctl+0x240/0xb48 > [<26aab344>] ret_fast_syscall+0x0/0x44 > > drm_atomic_helper_setup_commit() acquires a reference to a drm_crtc_commit > for each CRTC and associated connectors and planes involved in a modeset. > 64-byte leaks map well to the size of a drm_crtc_commit on 32-bit > architectures, and the second 4-byte chunk in the hex dump above awfully > looks like a reference count. > > We tracked this missing reference decrement down to the driver's > atmel_hlcdc_plane_atomic_destroy_state() callback. Its CRTC counterpart, > atmel_hlcdc_crtc_destroy_state(), calls into the drm_atomic helpers and > properly releases its references to the commit. Planes didn't. Using the > default helper for that purpose, __drm_atomic_helper_plane_destroy_state(), > fixes the leak and avoids reimplementing the same logic in the driver. > > [0]: https://gitlab.freedesktop.org/mesa/drm/-/tree/main/tests/modetest > Invoke with `modetest -M atmel-hlcdc -s 32:#0 -v`, assuming 32 is the > ID of a connector. > > Signed-off-by: Arnaud Lahache <a.lahache@xxxxxxxxxx> > Co-developed-by: Pierre-Louis Dourneau <pl.dourneau@xxxxxxxxxx> > Signed-off-by: Pierre-Louis Dourneau <pl.dourneau@xxxxxxxxxx> > Co-developed-by: Benoît Alcaina <b.alcaina@xxxxxxxxxx> > Signed-off-by: Benoît Alcaina <b.alcaina@xxxxxxxxxx> > --- > As far as our testing goes, we've been running 6 of our production units > with this patch for more than 2 weeks as per the date this patch is sent > out. We can report stable memory usage. Hello Arnaud, This patch fixes the memory leak but introduces a crash on my side when exiting a graphics app using the Microchip graphics library. ------------[ cut here ]------------ WARNING: CPU: 0 PID: 201 at lib/refcount.c:28 refcount_warn_saturate+0x58/0x130 refcount_t: underflow; use-after-free. Modules linked in: CPU: 0 PID: 201 Comm: basic Not tainted 6.1.55-linux4microchip-2023.10+ #65 Hardware name: Microchip SAM9X60 unwind_backtrace from show_stack+0x10/0x18 show_stack from dump_stack_lvl+0x28/0x34 dump_stack_lvl from __warn+0x70/0xb8 __warn from warn_slowpath_fmt+0x78/0xac warn_slowpath_fmt from refcount_warn_saturate+0x58/0x130 refcount_warn_saturate from kref_put+0x54/0x5c kref_put from drm_fb_release+0x100/0x11c drm_fb_release from drm_file_free+0xcc/0x1bc drm_file_free from drm_release+0x44/0x94 drm_release from __fput+0xf0/0x20c __fput from task_work_run+0x8c/0xb0 task_work_run from do_exit+0x310/0x760 do_exit from sys_exit_group+0x0/0x14 sys_exit_group from get_signal+0x524/0x64c get_signal from do_work_pending+0xf4/0x398 do_work_pending from slow_work_pending+0xc/0x24 Exception stack(0xc9991fb0 to 0xc9991ff8) 1fa0: 0054dd48 00000000 005331f8 00000001 1fc0: 00533270 00000002 00000000 404c8000 beb54170 0051b448 43ab8000 0051b424 1fe0: b6cc1d98 beb54040 b6c224fc b68357dc 20000010 ffffffff ---[ end trace 0000000000000000 ]--- 8<--- cut here --- Unable to handle kernel NULL pointer dereference at virtual address 00000004 [00000004] *pgd=00000000 Internal error: Oops: 805 [#1] ARM Modules linked in: CPU: 0 PID: 201 Comm: basic Tainted: G W 6.1.55-linux4microchip-2023.10+ #65 Hardware name: Microchip SAM9X60 PC is at drm_fb_release+0xc0/0x11c LR is at drm_fb_release+0x100/0x11c pc : [<c0329b10>] lr : [<c0329b50>] psr: 80000013 sp : c9991e48 ip : 00000000 fp : 0051b424 r10: c174a5f0 r9 : 00000000 r8 : c2188074 r7 : 60000013 r6 : c9991e5c r5 : ffffff8c r4 : c2188048 r3 : c1590994 r2 : c2188048 r1 : 00000000 r0 : c1590920 Flags: Nzcv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none Control: 0005317f Table: 2384c000 DAC: 00000051 Register r0 information: slab kmalloc-192 start c1590920 pointer offset 0 size 192 Register r1 information: NULL pointer Register r2 information: slab kmalloc-192 start c2188000 pointer offset 72 size 192 Register r3 information: slab kmalloc-192 start c1590920 pointer offset 116 size 192 Register r4 information: slab kmalloc-192 start c2188000 pointer offset 72 size 192 Register r5 information: non-paged memory Register r6 information: 2-page vmalloc region starting at 0xc9990000 allocated at kernel_clone+0xb4/0x204 Register r7 information: non-paged memory Register r8 information: slab kmalloc-192 start c2188000 pointer offset 116 size 192 Register r9 information: NULL pointer Register r10 information: slab task_struct start c174a140 pointer offset 1200 Register r11 information: non-paged memory Register r12 information: NULL pointer Process basic (pid: 201, stack limit = 0x41541c7b) Stack: (0xc9991e48 to 0xc9992000) 1e40: 00000001 00000000 00000000 00000000 00000000 c9991e5c 1e60: c9991e5c 5c44a9dd c2188000 c1621800 c2188060 c03146e4 000000c9 0000e200 1e80: 00000001 00000000 00000000 c1621800 c38f2cc0 c0fb63e0 c0c1ce70 c0f99560 1ea0: 00000000 c0314a5c c38f2cc0 c10ee6e8 000a201f c00c66a0 c38f2cc0 00000001 1ec0: c09f1a4c c00c67fc c38f2e00 c174a140 c0a20b1c c2232a50 c9991ef4 c002f000 1ee0: c174a140 c2232a20 c174a140 c001a094 00000002 00000008 c383c880 5c44a9dd 1f00: 00000002 c383c880 0051b424 c001a6a8 00000009 c0024af0 beb53c50 00000000 1f20: 00000000 5c44a9dd 00000000 00000000 c9991fb0 c174a140 00000000 00000000 1f40: 00000000 00000000 0051b424 c000c244 00000000 00000000 00000000 00000000 1f60: 00000000 00000000 00000009 00000000 00000000 00000000 00000000 00000000 1f80: 00000000 00000000 00000000 5c44a9dd b68357dc 20000010 ffffffff 0005317f 1fa0: 00000000 c174a140 43ab8000 c00084dc 0054dd48 00000000 005331f8 00000001 1fc0: 00533270 00000002 00000000 404c8000 beb54170 0051b448 43ab8000 0051b424 1fe0: b6cc1d98 beb54040 b6c224fc b68357dc 20000010 ffffffff 00000000 00000000 drm_fb_release from drm_file_free+0xcc/0x1bc drm_file_free from drm_release+0x44/0x94 drm_release from __fput+0xf0/0x20c __fput from task_work_run+0x8c/0xb0 task_work_run from do_exit+0x310/0x760 do_exit from sys_exit_group+0x0/0x14 sys_exit_group from get_signal+0x524/0x64c get_signal from do_work_pending+0xf4/0x398 do_work_pending from slow_work_pending+0xc/0x24 Exception stack(0xc9991fb0 to 0xc9991ff8) 1fa0: 0054dd48 00000000 005331f8 00000001 1fc0: 00533270 00000002 00000000 404c8000 beb54170 0051b448 43ab8000 0051b424 1fe0: b6cc1d98 beb54040 b6c224fc b68357dc 20000010 ffffffff Code: e590c018 e5902078 e5901074 e35c0001 (e5812004) ---[ end trace 00000000c0a20288 ]--- Fixing recursive fault but reboot is needed! The memory leak had been introduced with this commit: commit eec44d44a3d2d00c8f663f13555d3dd280b1ea3f Author: Daniel Vetter <daniel.vetter@xxxxxxxx> Date: Thu Jan 21 16:29:54 2021 +0100 drm/atmel: Use drm_atomic_helper_commit One of these drivers that predates the nonblocking support in helpers, and hand-rolled its own thing. Entirely not anything specific here, we can just delete it all and replace it with the helper version. Could also perhaps use the drm_mode_config_helper_suspend/resume stuff, for another few lines deleted. But I'm not looking at that stuff, I'm just going through all the atomic commit functions and make sure they have properly annotated dma-fence critical sections everywhere. I think this move to the drm atomic helper should have gone with an update on the release side as well. There is probably something wrong with the atomic_destroy_state callbacks provided by the atmel-hlcdc driver. Regards, Ludovic > > Admittedly, our usage of the DRM uAPI is rather simple: create 2 dumb > buffers, do an initial MODE_SETCRTC, and then MODE_PAGE_FLIP between the > two dumb buffers at the rate of once per second on average. We haven't > evaluated more complex workloads. > > drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c | 3 +-- > 1 file changed, 1 insertion(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c > index daa508504f47..390c4fc62af7 100644 > --- a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c > +++ b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_plane.c > @@ -934,8 +934,7 @@ static void atmel_hlcdc_plane_atomic_destroy_state(struct drm_plane *p, > state->dscrs[i]->self); > } > > - if (s->fb) > - drm_framebuffer_put(s->fb); > + __drm_atomic_helper_plane_destroy_state(s); > > kfree(state); > } > > base-commit: 9dfc46c87cdc8f5a42a71de247a744a6b8188980 > -- > 2.34.1 >
