fix race condition and NULL ptr deref in PCIe threaded probe

[AlomeProg]

Description

This PR adds a patch to fix a critical kernel panic in the Rockchip PCIe driver (pcie-dw-rockchip.c) occurring on threaded initialization failure.

Documentation summary for feature / change

When CONFIG_PCIE_RK_THREADED_INIT is enabled, a race condition exists between the probe and remove paths. If the hardware initialization fails (e.g., power supply issues or PHY errors), rk_pcie_remove may execute concurrently with the error handling path in the probe thread.

The original implementation had multiple flaws:

1. Unsafe Synchronization: It used a manual while loop with schedule_timeout instead of proper kernel synchronization primitives.
2. Broken Timeout Logic: The loop condition time_before(start, start + timeout) was always true (due to start not being updated), potentially leading to infinite loops.
3. NULL Pointer Dereference: The finish_probe flag was set to true even during error cleanup. This caused rk_pcie_remove to attempt accessing uninitialized structures (rk_pcie->pci->pp.bridge->bus), resulting in a Kernel Oops (Unable to handle kernel NULL pointer dereference at virtual address 00000000000002f8).

My solution:

The patch refactors the synchronization mechanism using standard kernel APIs:

1. Replaces the busy-wait loop with struct completion (probe_done) to ensure safe thread synchronization.
2. Replaces the ambiguous finish_probe flag with probe_ok to explicitly signal successful initialization.
3. Adds a check for probe_ok in rk_pcie_remove to exit early if initialization failed, preventing access to invalid pointers.
4. Ensures completion is initialized immediately after memory allocation to prevent usage of uninitialized data.

Testing

[x] Tested on RK3588 platform.
[x] Verified that the kernel boots successfully.
[x] Verified that no regression is introduced for successful probe scenarios.

dmesg log

[   27.170851] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   27.171511] pc : rk_pcie_remove+0x94/0x1bc
[   27.171915] lr : platform_remove+0x5c/0x74
[   27.172315] sp : ffff80000ccdbd50
[   27.172508] rk-pcie fe160000.pcie: can't get current limit.
[   27.172634] x29: ffff80000ccdbd50
[   27.173007] rk-pcie fe160000.pcie: host bridge /pcie@fe160000 ranges:
[   27.173161]  x28: 0000000000000000
[   27.173209] rk-pcie fe160000.pcie:       IO 0x00f1100000..0x00f11fffff -> 0x00f1100000
[   27.173231]  x27: 0000000000000000
[   27.173265] rk-pcie fe160000.pcie:      MEM 0x00f1200000..0x00f1ffffff -> 0x00f1200000
[   27.173306] 
[   27.173336] rk-pcie fe160000.pcie:      MEM 0x0940000000..0x097fffffff -> 0x0940000000
[   27.173382] x26: 0000000000000000
[   27.173449] rk-pcie fe160000.pcie: iATU unroll: enabled
[   27.173739]  x25: 00000000ffffffed
[   27.173764] rk-pcie fe160000.pcie: iATU regions: 8 ob, 8 ib, align 64K, limit 8G
[   27.173799]  x24: ffff00010149c000
[   27.179307] x23: 0000000000000001 x22: 0000000000000001 x21: ffff000107ee0000
[   27.179983] x20: ffff00010149c010 x19: ffff000100b04880 x18: 0000000000000000
[   27.180659] x17: 00656963702e3030 x16: 3030353165663a6d x15: 726f6674616c702b
[   27.181335] x14: 0000000000000000 x13: 64656c696166206f x12: 0000000000000000
[   27.182011] x11: ffff000100753120 x10: ffff0001003f7a10 x9 : ffff80000899f028
[   27.182687] x8 : 0101010101010101 x7 : 7f7f7f7f7f7f7f7f x6 : 736d47ff6364626d
[   27.183363] x5 : 0000008000000000 x4 : 0000000000000006 x3 : 000000000000000c
[   27.184038] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff000101b7a480
[   27.184715] Call trace:
[   27.184958]  rk_pcie_remove+0x94/0x1bc
[   27.185321]  platform_remove+0x5c/0x74
[   27.185681]  device_remove+0x54/0x7c
[   27.186039]  device_release_driver_internal+0x94/0x150
[   27.186535]  device_release_driver+0x20/0x2c
[   27.186949]  rk_pcie_really_probe+0x8f0/0x918
[   27.187373]  kthread+0xc4/0xd4
[   27.187684]  ret_from_fork+0x10/0x20

[   27.159256] Unable to handle kernel NULL pointer dereference at virtual address 00000000000002f8
[   27.159263] Mem abort info:
[   27.159266]   ESR = 0x0000000096000004
[   27.159270]   EC = 0x25: DABT (current EL), IL = 32 bits
[   27.159275]   SET = 0, FnV = 0
[   27.159279]   EA = 0, S1PTW = 0
[   27.159283]   FSC = 0x04: level 0 translation fault
[   27.159288] Data abort info:
[   27.159291]   ISV = 0, ISS = 0x00000004
[   27.159295]   CM = 0, WnR = 0
[   27.159299] user pgtable: 4k pages, 48-bit VAs, pgdp=0000000109c4d000
[   27.159305] [00000000000002f8] pgd=0000000000000000, p4d=0000000000000000
[   27.159319] Internal error: Oops: 0000000096000004 [#1] SMP