Created attachment 228433 [details] proposed patch Setting MTU `ifconfig mlxen0 mtu 9000` leads to panic: Unread portion of the kernel message buffer: panic: m_getzone: invalid cluster size 1522 cpuid = 14 time = 1630422868 KDB: stack backtrace: db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xfffffe01500e5710 vpanic() at vpanic+0x182/frame 0xfffffe01500e5760 panic() at panic+0x43/frame 0xfffffe01500e57c0 debugnet_mbuf_reinit() at debugnet_mbuf_reinit+0x22b/frame 0xfffffe01500e5800 debugnet_any_ifnet_update() at debugnet_any_ifnet_update+0x14b/frame 0xfffffe01500e5850 ifhwioctl() at ifhwioctl+0x48e/frame 0xfffffe01500e58d0 ifioctl() at ifioctl+0x3ac/frame 0xfffffe01500e59a0 mlx4_en_debugnet_init() uses priv->rx_mb_size to set mbuf cluster size in DEBUGNET. It is calculated as: priv->rx_mb_size = dev->if_mtu + ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN; But DEBUGNET uses this value to call debugnet_mbuf_reinit(), where m_getzone() is called for specified cluster size. And since it has wrong value, this triggers panic.
I'm good with your patch. Just wait for kib@ to approve aswell. Sorry for not getting back to you on this one. Has been very hectic. --HPS
You can just submit and MFC it once Konstantin approves it.
OK, got a green light from Konstantin. Please go ahead and commit and MFC. --HPS
A commit in branch main references this bug: URL: https://cgit.FreeBSD.org/src/commit/?id=5a7de2b42caf6241e87b417a0521e9ab303989d7 commit 5a7de2b42caf6241e87b417a0521e9ab303989d7 Author: Hans Petter Selasky <hselasky@FreeBSD.org> AuthorDate: 2021-10-05 08:46:56 +0000 Commit: Hans Petter Selasky <hselasky@FreeBSD.org> CommitDate: 2021-10-05 08:48:30 +0000 mlx4en(4): Fix wrong mbuf cluster size in mlx4_en_debugnet_init() This fixes an "invalid cluster size" panic when debugnet is activated. panic() m_getzone() debugnet_mbuf_reinit() debugnet_any_ifnet_update() ifhwioctl() ifioctl() Submitted by: ae@ PR: 258923 MFC after: 1 week Sponsored by: NVIDIA Networking sys/dev/mlx4/mlx4_en/mlx4_en_netdev.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
A commit in branch stable/13 references this bug: URL: https://cgit.FreeBSD.org/src/commit/?id=fe5ee07a11ced27eab9c6a6bd0f3e27d99423ba8 commit fe5ee07a11ced27eab9c6a6bd0f3e27d99423ba8 Author: Hans Petter Selasky <hselasky@FreeBSD.org> AuthorDate: 2021-10-05 08:46:56 +0000 Commit: Hans Petter Selasky <hselasky@FreeBSD.org> CommitDate: 2021-10-12 12:12:00 +0000 mlx4en(4): Fix wrong mbuf cluster size in mlx4_en_debugnet_init() This fixes an "invalid cluster size" panic when debugnet is activated. panic() m_getzone() debugnet_mbuf_reinit() debugnet_any_ifnet_update() ifhwioctl() ifioctl() Submitted by: ae@ PR: 258923 Sponsored by: NVIDIA Networking (cherry picked from commit 5a7de2b42caf6241e87b417a0521e9ab303989d7) sys/dev/mlx4/mlx4_en/mlx4_en_netdev.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
A commit in branch stable/12 references this bug: URL: https://cgit.FreeBSD.org/src/commit/?id=8d804809630fd145683de92ba502a9ed34b94115 commit 8d804809630fd145683de92ba502a9ed34b94115 Author: Hans Petter Selasky <hselasky@FreeBSD.org> AuthorDate: 2021-10-05 08:46:56 +0000 Commit: Hans Petter Selasky <hselasky@FreeBSD.org> CommitDate: 2021-10-12 12:17:02 +0000 mlx4en(4): Fix wrong mbuf cluster size in mlx4_en_debugnet_init() This fixes an "invalid cluster size" panic when debugnet is activated. panic() m_getzone() debugnet_mbuf_reinit() debugnet_any_ifnet_update() ifhwioctl() ifioctl() Submitted by: ae@ PR: 258923 Sponsored by: NVIDIA Networking (cherry picked from commit 5a7de2b42caf6241e87b417a0521e9ab303989d7) sys/dev/mlx4/mlx4_en/mlx4_en_netdev.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
is this considered "fixed" by now? This morning I swapped (ok, I TRYED TO SWAP) an Intel 10Gbe card to a Mellanox Connect-X3. At the next boot this error showed here up too, no way to get into multiuser. Architecture: amd64 Architecture Version: 2 Dump Length: 1162113024 Blocksize: 512 Compression: none Dumptime: 2022-02-20 07:47:05 +0100 Hostname: l3router.meiszl.de Magic: FreeBSD Kernel Dump Version String: FreeBSD 13.0-RELEASE-p7 #0: Mon Jan 31 18:24:03 UTC 2022 root@amd64-builder.daemonology.net:/usr/obj/usr/src/amd64.amd64/sys/GENERIC Panic String: m_getzone: invalid cluster size 1522 Dump Parity: 3248958317 Bounds: 4 Dump Status: good I've tried a few times, disabled ipfw (had some troubles with it last year but this has been fixed now), but finally had the cards flipped back again to get the box up and running. So, if the fix here has already been incorporated, it did not work, at least for me :-(
It has not been merged to 13.0, you need to use 13-stable!
(In reply to Hans Petter Selasky from comment #8) >It has not been merged to 13.0, you need to use 13-stable! Sorry, I dont get it. Why should I use 13-stable if it is not in 13.0 ??? I've added it manually today and created a custom kernel, now it works. But I dont understand why it is not incorporated into the normal updates that come in regulary ? Usually I avoid using "special" kernels.
It will be fixed in the upcoming 13.1-RELEASE. Please compare: https://docs.freebsd.org/en/books/dev-model/#development-model