CVE-2022-50045
Severity CVSS v4.0:
Pending analysis
Type:
Unavailable / Other
Publication date:
18/06/2025
Last modified:
18/06/2025
Description
In the Linux kernel, the following vulnerability has been resolved:<br />
<br />
powerpc/pci: Fix get_phb_number() locking<br />
<br />
The recent change to get_phb_number() causes a DEBUG_ATOMIC_SLEEP<br />
warning on some systems:<br />
<br />
BUG: sleeping function called from invalid context at kernel/locking/mutex.c:580<br />
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 1, name: swapper<br />
preempt_count: 1, expected: 0<br />
RCU nest depth: 0, expected: 0<br />
1 lock held by swapper/1:<br />
#0: c157efb0 (hose_spinlock){+.+.}-{2:2}, at: pcibios_alloc_controller+0x64/0x220<br />
Preemption disabled at:<br />
[] 0x0<br />
CPU: 0 PID: 1 Comm: swapper Not tainted 5.19.0-yocto-standard+ #1<br />
Call Trace:<br />
[d101dc90] [c073b264] dump_stack_lvl+0x50/0x8c (unreliable)<br />
[d101dcb0] [c0093b70] __might_resched+0x258/0x2a8<br />
[d101dcd0] [c0d3e634] __mutex_lock+0x6c/0x6ec<br />
[d101dd50] [c0a84174] of_alias_get_id+0x50/0xf4<br />
[d101dd80] [c002ec78] pcibios_alloc_controller+0x1b8/0x220<br />
[d101ddd0] [c140c9dc] pmac_pci_init+0x198/0x784<br />
[d101de50] [c140852c] discover_phbs+0x30/0x4c<br />
[d101de60] [c0007fd4] do_one_initcall+0x94/0x344<br />
[d101ded0] [c1403b40] kernel_init_freeable+0x1a8/0x22c<br />
[d101df10] [c00086e0] kernel_init+0x34/0x160<br />
[d101df30] [c001b334] ret_from_kernel_thread+0x5c/0x64<br />
<br />
This is because pcibios_alloc_controller() holds hose_spinlock but<br />
of_alias_get_id() takes of_mutex which can sleep.<br />
<br />
The hose_spinlock protects the phb_bitmap, and also the hose_list, but<br />
it doesn&#39;t need to be held while get_phb_number() calls the OF routines,<br />
because those are only looking up information in the device tree.<br />
<br />
So fix it by having get_phb_number() take the hose_spinlock itself, only<br />
where required, and then dropping the lock before returning.<br />
pcibios_alloc_controller() then needs to take the lock again before the<br />
list_add() but that&#39;s safe, the order of the list is not important.
Impact
References to Advisories, Solutions, and Tools
- https://git.kernel.org/stable/c/1d9e75c3d8cdf7c96a94cb77450d4ee070279e6a
- https://git.kernel.org/stable/c/5db5ce0f1963c6c8275719a80cb65e9c98d32726
- https://git.kernel.org/stable/c/6f75057c21eab12c6ccb7f06f859641a6edfab99
- https://git.kernel.org/stable/c/8d48562a2729742f767b0fdd994d6b2a56a49c63
- https://git.kernel.org/stable/c/90f195c01a2e8d8da6281791617e21109719c981
- https://git.kernel.org/stable/c/a868f771ee41c97a25a04b8c632a7f06689b307b
- https://git.kernel.org/stable/c/ccb0a42d3f40c436295e0fef57ab613ae5b925a4