[REGRESSION, BISECTED] IPv6 RA is broken with Linux 6.12.42+

[wangzijie]

> Hi, everyone,
> 
> 
> We noticed a regression in OpenWrt, with IPv6, which causes a router's
> client devices to stop receiving the IPv6 default route. I have
> bisected it down to (rather surprisingly)
> fc1072d934f687e1221d685cf1a49a5068318f34 ("proc: use the same
> treatment to check proc_lseek as ones for proc_read_iter et.al").
> Reverting the aforementioned commit fixes the issue, of course.
> 
> Git bisect log follows:
> 
> git bisect start
> # status: waiting for both good and bad commits
> # bad: [880e4ff5d6c8dc6b660f163a0e9b68b898cc6310] Linux 6.12.42
> git bisect bad 880e4ff5d6c8dc6b660f163a0e9b68b898cc6310
> # status: waiting for good commit(s), bad commit known
> # good: [8f5ff9784f3262e6e85c68d86f8b7931827f2983] Linux 6.12.41
> git bisect good 8f5ff9784f3262e6e85c68d86f8b7931827f2983
> # good: [dab173bae3303f074f063750a8dead2550d8c782] RDMA/hns: Fix
> double destruction of rsv_qp
> git bisect good dab173bae3303f074f063750a8dead2550d8c782
> # bad: [11fa01706a4f60e759fbee7c53095ff22eaf1595] PCI: pnv_php: Work
> around switches with broken presence detection
> git bisect bad 11fa01706a4f60e759fbee7c53095ff22eaf1595
> # bad: [966460bace9e1dd8609c9d44cf4509844daea8bb] perf record: Cache
> build-ID of hit DSOs only
> git bisect bad 966460bace9e1dd8609c9d44cf4509844daea8bb
> # bad: [f63bd615e58f43dbe4b2e4c3f3ffa0bfb7766007] hwrng: mtk - handle
> devm_pm_runtime_enable errors
> git bisect bad f63bd615e58f43dbe4b2e4c3f3ffa0bfb7766007
> # bad: [9ea3f6b9a67be3476e331ce51cac316c2614a564] pinmux: fix race
> causing mux_owner NULL with active mux_usecount
> git bisect bad 9ea3f6b9a67be3476e331ce51cac316c2614a564
> # good: [1209e33fe3afb6d9e543f963d41b30cfc04538ff] RDMA/hns: Get
> message length of ack_req from FW
> git bisect good 1209e33fe3afb6d9e543f963d41b30cfc04538ff
> # good: [5f3c0301540bc27e74abbfbe31571e017957251b] RDMA/hns: Fix
> -Wframe-larger-than issue
> git bisect good 5f3c0301540bc27e74abbfbe31571e017957251b
> # bad: [fc1072d934f687e1221d685cf1a49a5068318f34] proc: use the same
> treatment to check proc_lseek as ones for proc_read_iter et.al
> git bisect bad fc1072d934f687e1221d685cf1a49a5068318f34
> # good: [ec437d0159681bbdb1cf1f26759d12e9650bffca] kernel: trace:
> preemptirq_delay_test: use offstack cpu mask
> git bisect good ec437d0159681bbdb1cf1f26759d12e9650bffca
> # first bad commit: [fc1072d934f687e1221d685cf1a49a5068318f34] proc:
> use the same treatment to check proc_lseek as ones for proc_read_iter
> et.al
> 
> Please let me know if you need any additional information.
> 
> 
> Kind regards,
> 
> Rui Salvaterra

Hi Rui,
Thanks. I have submitted a patch and I think it can fix this.
https://lore.kernel.org/all/20250821105806.1453833-1-wangzijie1@honor.com