Error UBSAN: array-index-out-of-bounds amdgpu (drivers/gpu/drm/amd/amdgpu/../pm/powerplay/hwmgr/smu7_hwmgr.c)

[msr1985]

Error in boot:

[ 8.597520] UBSAN: array-index-out-of-bounds in /build/linux-D15vQj/linux-6.5.0/drivers/gpu/drm/amd/amdgpu/../pm/powerplay/hwmgr/smu7_hwmgr.c:3676:4
[ 8.597527] index 7 is out of range for type 'ATOM_Polaris_SCLK_Dependency_Record [1]'

ProblemType: Bug
DistroRelease: Ubuntu 23.10
Package: linux-image-generic 6.5.0.9.11
ProcVersionSignature: Ubuntu 6.5.0-9.9-generic 6.5.3
Uname: Linux 6.5.0-9-generic x86_64
ApportVersion: 2.27.0-0ubuntu5
Architecture: amd64
CasperMD5CheckResult: pass
CurrentDesktop: ubuntu:GNOME
Date: Fri Oct 20 09:28:16 2023
InstallationDate: Installed on 2022-10-12 (373 days ago)
InstallationMedia: Ubuntu 22.04.1 LTS "Jammy Jellyfish" - Release amd64 (20220809.1)
MachineType: {report['dmi.sys.vendor']} {report['dmi.product.name']}
ProcFB: 0 amdgpudrmfb
ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-6.5.0-9-generic root=UUID=9edc5478-c6c2-4cf3-9de8-01ccb697fb9e ro quiet splash audit=0 mitigations=off amdgpu.ppfeaturemask=0xffffffff vt.global_cursor_default=0 loglevel=2 rd.systemd.show_status=false rd.udev.log-prority=3 sysrq_always_enabled=1 audit=0 vt.handoff=7
RelatedPackageVersions:
 linux-restricted-modules-6.5.0-9-generic N/A
 linux-backports-modules-6.5.0-9-generic N/A
 linux-firmware 20230919.git3672ccab-0ubuntu2.1
SourcePackage: linux
UpgradeStatus: Upgraded to mantic on 2023-10-13 (6 days ago)
dmi.bios.date: 07/11/2014
dmi.bios.release: 4.6
dmi.bios.vendor: American Megatrends Inc.
dmi.bios.version: D3EMW08.110
dmi.board.asset.tag: To be filled by O.E.M.
dmi.board.name: D3F3-EM
dmi.board.vendor: MEDION
dmi.board.version: 1.0
dmi.chassis.type: 3
dmi.chassis.vendor: MEDION
dmi.modalias: dmi:bvnAmericanMegatrendsInc.:bvrD3EMW08.110:bd07/11/2014:br4.6:svnMEDION:pnD3F3-EM:pvr1.0:rvnMEDION:rnD3F3-EM:rvr1.0:cvnMEDION:ct3:cvr:skuTobefilledbyO.E.M.:
dmi.product.family: To be filled by O.E.M.
dmi.product.name: D3F3-EM
dmi.product.sku: To be filled by O.E.M.
dmi.product.version: 1.0
dmi.sys.vendor: MEDION

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This change was made by a bot.

[Mario Limonciello (superm1)]

Also reported at #2036742

It's caused by the use of flexible size arrays.
Here's the fixes from linux-next that should help it.

https://cgit.freedesktop.org/drm/drm/commit/drivers/gpu/drm/amd/pm/powerplay/hwmgr?id=0f0e59075b5c22f1e871fbd508d6e4f495048356
https://cgit.freedesktop.org/drm/drm/commit/drivers/gpu/drm/amd/pm/powerplay/hwmgr?id=760efbca74a405dc439a013a5efaa9fadc95a8c3

[Olivier Duclos (odc)]

The mentioned patches fix the issue for Polaris and Tonga, but other AMD GPUs are also affected. I have a Radeon Vega 64 and here is a summary of the UBSAN errors I see:

index 2 is out of range for type 'ATOM_Vega10_MM_Dependency_Record [1]'
index 1 is out of range for type 'ATOM_Vega10_CLK_Dependency_Record [1]'
index 5 is out of range for type 'ATOM_Vega10_CLK_Dependency_Record [1]'
index 1 is out of range for type 'ATOM_Vega10_MCLK_Dependency_Record [1]'
index 1 is out of range for type 'ATOM_Vega10_PCIE_Record [1]'
index 1 is out of range for type 'ATOM_Vega10_Voltage_Lookup_Record [1]'

[Alex Deucher (alexander-deucher)]

FWIW, none of these are actually out of bound accesses. These just happen to use the old nomenclature for variable sized arrays.

[Alex Deucher (alexander-deucher)]

This updates the rest of them:
https://patchwork.freedesktop.org/patch/564786/

[misiu_mp (misiu-mp)]

I have HAWAII (R9 290X). I get the following:

array-index-out-of-bounds in /build/linux-7T0Dsf/linux-6.6.6/drivers/gpu/drm/radeon/radeon_atombios.c:2717:34
index 18 is out of range for type 'UCHAR [1]'

array-index-out-of-bounds in /build/linux-7T0Dsf/linux-6.6.6/drivers/gpu/drm/radeon/radeon_atombios.c:2715:55
index 1 is out of range for type 'UCHAR [1]'

UBSAN: array-index-out-of-bounds in /build/linux-7T0Dsf/linux-6.6.6/drivers/gpu/drm/radeon/ci_dpm.c:5588:32
index 9 is out of range for type 'UCHAR [1]'

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Muhammed Sabbagh (muhammedsabbagh)]

similar to other reports from the fellows up:

GPU: \ Vendor: AMD (0x1002)
    Device: AMD Radeon R7 M340 (iceland, LLVM 15.0.7, DRM 3.54, 6.5.0-14-generic) (0x6900)
    Version: 23.0.4

Linux: Ubuntu 22.04
Kernel: linux-image-6.5.0-14-generic/jammy-updates,jammy-security,now 6.5.0-14.14~22.04.1 amd64

[ 12.308765] ================================================================================
[ 12.309790] ================================================================================
[ 12.310831] UBSAN: array-index-out-of-bounds in /build/linux-hwe-6.5-q7NZ0T/linux-hwe-6.5-6.5.0/drivers/gpu/drm/amd/amdgpu/../pm/powerplay/hwmgr/processpptables.c:1279:65
[ 12.312963] index 1 is out of range for type 'ATOM_PPLIB_SAMClk_Voltage_Limit_Record [1]'
[ 12.314267] CPU: 1 PID: 341 Comm: systemd-udevd Not tainted 6.5.0-14-generic #14~22.04.1-Ubuntu
[ 12.314275] Hardware name: LENOVO 80SY/LNVNB161216, BIOS 0ZCN52WW 12/11/2019
[ 12.314278] Call Trace:
[ 12.314281] <TASK>
[ 12.314286] dump_stack_lvl+0x48/0x70
[ 12.314298] dump_stack+0x10/0x20
[ 12.314304] __ubsan_handle_out_of_bounds+0xc6/0x110
[ 12.314314] init_clock_voltage_dependency+0x8ca/0xa60 [amdgpu]
[ 12.315266] pp_tables_initialize+0x116/0x440 [amdgpu]
[ 12.316037] ? amdgpu_ring_test_helper+0x83/0x90 [amdgpu]
[ 12.316822] hwmgr_hw_init+0x78/0x1e0 [amdgpu]
[ 12.317642] pp_hw_init+0x16/0x50 [amdgpu]
[ 12.318433] amdgpu_device_ip_init+0x48a/0x960 [amdgpu]
[ 12.319116] amdgpu_device_init+0x9c8/0x1160 [amdgpu]
[ 12.319779] amdgpu_driver_load_kms+0x1a/0x1c0 [amdgpu]
[ 12.320436] amdgpu_pci_probe+0x182/0x450 [amdgpu]
[ 12.321077] local_pci_probe+0x44/0xb0
[ 12.321087] pci_call_probe+0x55/0x190
[ 12.321092] pci_device_probe+0x84/0x120
[ 12.321098] really_probe+0x1c9/0x430
[ 12.321103] __driver_probe_device+0x8c/0x190
[ 12.321107] driver_probe_device+0x24/0xd0
[ 12.321111] __driver_attach+0x10b/0x210
[ 12.321115] ? __pfx___driver_attach+0x10/0x10
[ 12.321120] bus_for_each_dev+0x8a/0xf0
[ 12.321126] driver_attach+0x1e/0x30
[ 12.321132] bus_add_driver+0x127/0x240
[ 12.321139] driver_register+0x5e/0x130
[ 12.321144] ? __pfx_amdgpu_init+0x10/0x10 [amdgpu]
[ 12.321817] __pci_register_driver+0x62/0x70
[ 12.321824] amdgpu_init+0x69/0xff0 [amdgpu]
[ 12.322454] do_one_initcall+0x5b/0x340
[ 12.322464] do_init_module+0x68/0x260
[ 12.322471] load_module+0xb85/0xcd0
[ 12.322479] ? security_kernel_post_read_file+0x75/0x90
[ 12.322484] ? security_kernel_post_read_file+0x75/0x90
[ 12.322491] init_module_from_file+0x96/0x100
[ 12.322497] ? init_module_from_file+0x96/0x100
[ 12.322509] idempotent_init_module+0x11c/0x2b0
[ 12.322517] __x64_sys_finit_module+0x64/0xd0
[ 12.322523] do_syscall_64+0x58/0x90
[ 12.322529] ? do_syscall_64+0x67/0x90
[ 12.322534] ? sysvec_call_function+0x4b/0xd0
[ 12.322539] entry_SYSCALL_64_after_hwframe+0x6e/0xd8
[ 12.322545] RIP: 0033:0x7f7bf951e88d
[ 12.322569] Code: 5b 41 5c c3 66 0f 1f 84 00 00 00 00 00 f3 0f 1e fa 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 73 b5 0f 00 f7 d8 64 89 ...

[StoatWblr (stoatwblr)]

Same issue with VERDE - (atombios and si_dpm) in both 6.5 and 6.8 series kernels on both radeon and amdgpu drivers

Whatever this is isn't just 1 or 2 places in the driver (and it's been occurring for several years)

Ubuntu Noble boot (dual W4100 quad-head cards), radeon driver - the amdgpu messages are similar

[ 11.030090] UBSAN: array-index-out-of-bounds in /build/linux-zdc93w/linux-6.8.0/drivers/gpu/drm/radeon/radeon_atombios.c:2718:34
[ 11.030095] index 48 is out of range for type 'UCHAR [1]'

[ 11.031839] UBSAN: array-index-out-of-bounds in /build/linux-zdc93w/linux-6.8.0/drivers/gpu/drm/radeon/radeon_atombios.c:2716:55
[ 11.031843] index 1 is out of range for type 'UCHAR [1]'

[ 11.032818] UBSAN: array-index-out-of-bounds in /build/linux-zdc93w/linux-6.8.0/drivers/gpu/drm/radeon/radeon_atombios.c:2706:39
[ 11.032823] index 2 is out of range for type 'ATOM_PPLIB_NONCLOCK_INFO [1]'

[ 11.033887] UBSAN: array-index-out-of-bounds in /build/linux-zdc93w/linux-6.8.0/drivers/gpu/drm/radeon/si_dpm.c:6831:39
[ 11.033891] index 2 is out of range for type 'ATOM_PPLIB_NONCLOCK_INFO [1]'

[ 11.034910] UBSAN: array-index-out-of-bounds in /build/linux-zdc93w/linux-6.8.0/drivers/gpu/drm/radeon/si_dpm.c:6868:32
[ 11.034913] index 16 is out of range for type 'UCHAR [1]'

[Mario Limonciello (superm1)]

https://patchwork.freedesktop.org/patch/588305/

[Mario Limonciello (superm1)]

Variable sized arrays have fallen out of fashion.

There are various patches like that which fix this issue every time it crops up.

If you can still reproduce on the latest 6.9-rc kernels you should report a bug upstream to get the remaining cases fixed. Or if you feel comfortable you can write a patch for any remaining cases you hit. As you can see it's a trivial fix.