BUG: Kernel NULL pointer dereference on read at 0x00000000 in pnv_get_random_long()

[Madhavan Srinivasan]

On 10/2/24 5:31 PM, Corentin LABBE wrote:
> Hello
> 
> I have a 8335-GCA POWER8 which got a kernel crash during boot:
> [   11.754238] Kernel attempted to read user page (0) - exploit attempt? (uid: 0)
> [   11.754437] BUG: Kernel NULL pointer dereference on read at 0x00000000
> [   11.754499] Faulting instruction address: 0xc0000000000c3758
> [   11.754518] Oops: Kernel access of bad area, sig: 11 [#1]
> [   11.754534] LE PAGE_SIZE=64K MMU=Hash SMP NR_CPUS=2048 NUMA PowerNV
> [   11.754699] Modules linked in: powernv_rng(+) ecb ctr sr_mod hid ofpart fb_sys_fops cdrom i2c_algo_bit powernv_flash sg mtd vmx_crypto(+) ipmi_powernv ipmi_devintf at24(+) ipmi_msghandler opal_prd regmap_i2c nfsd gf128mul auth_rpcgss nfs_acl lockd grace sunrpc drm fuse configfs loop drm_panel_orientation_quirks ip_tables x_tables autofs4 uas usb_storage ext4 crc16 mbcache jbd2 crc32c_generic dm_mod xhci_pci xhci_hcd sd_mod t10_pi crc64_rocksoft crc64 crc_t10dif crct10dif_generic crct10dif_common usbcore tg3 libphy crc32c_vpmsum ahci usb_common libahci
> [   11.754869] CPU: 25 PID: 1332 Comm: (udev-worker) Not tainted 6.1.106 #4 
> [   11.754890] Hardware name: 8335-GCA POWER8 (raw) 0x4d0200 opal:skiboot-5.4.8-5787ad3 PowerNV
> [   11.754926] NIP:  c0000000000c3758 LR: c0000000000c3754 CTR: 0000000000000000
> [   11.754947] REGS: c00000000ec3af70 TRAP: 0300   Not tainted  (6.1.106)
> [   11.754966] MSR:  900000000280b033 <SF,HV,VEC,VSX,EE,FP,ME,IR,DR,RI,LE>  CR: 44222282  XER: 20000000
> [   11.755168] CFAR: c0000000001dfbb4 DAR: 0000000000000000 DSISR: 40000000 IRQMASK: 0 
>                GPR00: c0000000000c3754 c00000000ec3b210 c00000000113cd00 000000000000002c 
>                GPR04: 00000000ffff7fff c00000000ec3b010 c00000000ec3b008 0000000ff57e0000 
>                GPR08: 0000000000000027 c000000ff7907f98 0000000000000001 0000000000002200 
>                GPR12: 0000000000000000 c000000ffffeaf00 0000000000000020 0000000022000000 
>                GPR16: 0000000000000000 0000000000000000 0000000000000009 000000013c86f5d8 
>                GPR20: 0000000000000000 000001002cd75d90 0000000000000000 0000000000000005 
>                GPR24: 000001002cd794a0 000001002cd75d90 c00000000285e6fc c000000000f9e4a0 
>                GPR28: 0000000000000003 0000000000000004 0000000000000000 c0000010103ca180 
> [   11.755363] NIP [c0000000000c3758] pnv_get_random_long+0x88/0x170
> [   11.755386] LR [c0000000000c3754] pnv_get_random_long+0x84/0x170
> [   11.755407] Call Trace:
> [   11.755416] [c00000000ec3b210] [c0000000000c3754] pnv_get_random_long+0x84/0x170 (unreliable)
> [   11.755444] [c00000000ec3b280] [c008000021c50130] powernv_rng_read+0x98/0x120 [powernv_rng]
> [   11.755473] [c00000000ec3b300] [c00000000091ac88] add_early_randomness+0x88/0x150
> [   11.755577] [c00000000ec3b340] [c00000000091b2c4] hwrng_register+0x344/0x420
> [   11.755678] [c00000000ec3b3a0] [c00000000091b408] devm_hwrng_register+0x68/0xf0
> [   11.755703] [c00000000ec3b3e0] [c008000021c5003c] powernv_rng_probe+0x34/0x90 [powernv_rng]
> [   11.755728] [c00000000ec3b450] [c000000000949218] platform_probe+0x78/0x110
> [   11.755750] [c00000000ec3b4d0] [c0000000009442d8] really_probe+0x108/0x590
> [   11.755773] [c00000000ec3b560] [c000000000944814] __driver_probe_device+0xb4/0x230
> [   11.755799] [c00000000ec3b5e0] [c0000000009449e4] driver_probe_device+0x54/0x130
> [   11.755824] [c00000000ec3b620] [c0000000009456d8] __driver_attach+0x158/0x2b0
> [   11.755850] [c00000000ec3b6a0] [c000000000940764] bus_for_each_dev+0xb4/0x140
> [   11.755874] [c00000000ec3b700] [c000000000943734] driver_attach+0x34/0x50
> [   11.755896] [c00000000ec3b720] [c000000000942d88] bus_add_driver+0x218/0x300
> [   11.755921] [c00000000ec3b7b0] [c000000000946b84] driver_register+0xb4/0x1c0
> [   11.755947] [c00000000ec3b820] [c000000000948b98] __platform_driver_register+0x38/0x50
> [   11.755969] [c00000000ec3b840] [c008000021c501e8] powernv_rng_driver_init+0x30/0x4c [powernv_rng]
> [   11.755997] [c00000000ec3b860] [c0000000000121b0] do_one_initcall+0x80/0x320
> [   11.756020] [c00000000ec3b940] [c0000000002198bc] do_init_module+0x6c/0x290
> [   11.756042] [c00000000ec3b9c0] [c00000000021d118] __do_sys_finit_module+0xd8/0x190
> [   11.756066] [c00000000ec3baf0] [c00000000002b038] system_call_exception+0x138/0x260
> [   11.756091] [c00000000ec3be10] [c00000000000c654] system_call_common+0xf4/0x258
> [   11.756117] --- interrupt: c00 at 0x7fffaae9a9e4
> [   11.756134] NIP:  00007fffaae9a9e4 LR: 00007fffab110500 CTR: 0000000000000000
> [   11.756153] REGS: c00000000ec3be80 TRAP: 0c00   Not tainted  (6.1.106)
> [   11.762944] MSR:  900000000280f033 <SF,HV,VEC,VSX,EE,PR,FP,ME,IR,DR,RI,LE>  CR: 24222248  XER: 00000000
> [   11.765251] IRQMASK: 0 
>                GPR00: 0000000000000161 00007ffff4b57210 00007fffaafa6f00 0000000000000006 
>                GPR04: 00007fffab11be88 0000000000000000 0000000000000006 0000000000000000 
>                GPR08: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 
>                GPR12: 0000000000000000 00007fffab1fe240 0000000000000020 0000000022000000 
>                GPR16: 0000000000000000 0000000000000000 0000000000000009 000000013c86f5d8 
>                GPR20: 0000000000000000 000001002cd75d90 0000000000000000 0000000000000005 
>                GPR24: 000001002cd794a0 000001002cd75d90 0000000022000000 000001002cd32120 
>                GPR28: 00007fffab11be88 0000000000020000 0000000000000000 000001002cd75d90 
> [   11.773845] NIP [00007fffaae9a9e4] 0x7fffaae9a9e4
> [   11.774334] LR [00007fffab110500] 0x7fffab110500
> [   11.774347] --- interrupt: c00
> [   11.779698] Instruction dump:
> [   11.779711] e88952f8 38634198 3bde52f8 4811c439 60000000 e94d0030 3c62ffe4 386341c0 
> [   11.779739] 7fcaf02a 7fc4f378 4811c41d 60000000 <e93e0000> 7c0004ac e9490000 0c0a0000 
> [   11.779782] ---[ end trace 0000000000000000 ]---
> 
> This happen on stock debian 6.1.0-23-powerpc64le.

I am not able to recreate this in my setup. 
Have tried stable 6.1.106, 6.1.100 and also latest upstream with powernv_defconfig.
Can you share the config file. 

Maddy

> 
> I debugged a bit and the crash happen in arch/powerpc/platforms/powernv/rng.c in pnv_get_random_long()
> The call rng = get_cpu_var(pnv_rng) return null and so rng is dereferenced via ->regs just after.
> 
> I have no real idea on how to fix properly (appart adding a "!rng return 0" test)
> 
> Regards
>