"kernel NULL pointer dereference" on modprobe zfs

[posativ]

I'm using the gentoo hardened profile with linux 3.7.10 and after some new configuration flags to support libvirt/qemu, I'm no longer able to start zfs:

SPL: Loaded module v0.6.0-rc14
znvpair: module license 'CDDL' taints kernel.
Disabling lock debugging due to kernel taint
BUG: unable to handle kernel NULL pointer dereference at 0000000000000004
IP: [<ffffffff81803059>] _raw_spin_lock+0x9/0x30
PGD 0 
Oops: 0002 [#1] SMP 
Modules linked in: znvpair(PO+) spl(O) xt_TCPMSS ipt_rpfilter xt_statistic xt_LOG xt_time xt_connlimit xt_realm xt_addrtype xt_comment xt_recent xt_nat ipt_MASQUERADE ipt_ECN ipt_ah nf_nat_sip nf_nat_irc nf_nat_ftp xt_tcpmss xt_pkttype xt_owner xt_NFQUEUE xt_NFLOG xt_multiport xt_mark xt_mac xt_limit xt_length xt_iprange xt_helper xt_hashlimit xt_DSCP xt_dscp xt_dccp xt_connmark xt_CLASSIFY xt_AUDIT w83627ehf hwmon_vid xt_conntrack iptable_nat nf_nat_ipv4 nf_nat arc4 ath9k_htc coretemp ath9k ath9k_common ath9k_hw ath mac80211 cfg80211
CPU 1 
Pid: 3897, comm: modprobe Tainted: P           O 3.7.10-gentoo #4                  /S1200KP
RIP: 0010:[<ffffffff81803059>]  [<ffffffff81803059>] _raw_spin_lock+0x9/0x30
RSP: 0018:ffff8803f39e9e88  EFLAGS: 00010282
RAX: 0000000000000100 RBX: 0000000000000000 RCX: 0000000000000000
RDX: ffff88040d01e0b0 RSI: ffff8803f39e9ed8 RDI: 0000000000000004
RBP: ffff8803f39e9e88 R08: 0000000000000000 R09: ffff88040f00accc
R10: 8080808080808080 R11: 0000000000000000 R12: ffff8803f39e9ed8
R13: ffffffffa020a950 R14: 00007f8a088fd301 R15: 0000000000040000
FS:  00007f8a0832a700(0000) GS:ffff88041fb00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000004 CR3: 00000003f39d0000 CR4: 00000000000407e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process modprobe (pid: 3897, threadinfo ffff8803f39e8000, task ffff88040d01e0b0)
Stack:
 ffff8803f39e9eb8 ffffffffa01f919a 00007ffffffff000 0000000000000000
 00007ffffffff000 0000000000000000 ffff8803f39e9ef8 ffffffffa01f928c
 0000000000000001 ffffffff00000000 ffff88040d5b2520 ffff88040be69000
Call Trace:
 [<ffffffffa01f919a>] set_fs_pwd+0x1a/0x60 [spl]
 [<ffffffffa01f928c>] vn_set_pwd+0xac/0xc0 [spl]
 [<ffffffffa01f9740>] spl_setup+0x10/0x30 [spl]
 [<ffffffffa020a959>] init_module+0x9/0x10 [znvpair]
 [<ffffffff810001fa>] do_one_initcall+0x3a/0x160
 [<ffffffff810d74da>] sys_init_module+0x8a/0x1f0
 [<ffffffff8180a912>] system_call_fastpath+0x16/0x1b
Code: 38 c2 74 0f 66 0f 1f 44 00 00 f3 90 0f b6 03 38 c2 75 f7 48 83 c4 08 5b 5d c3 0f 1f 84 00 00 00 00 00 55 b8 00 01 00 00 48 89 e5 <f0> 66 0f c1 07 0f b6 d4 38 c2 74 0c 0f 1f 00 f3 90 0f b6 07 38 
RIP  [<ffffffff81803059>] _raw_spin_lock+0x9/0x30
 RSP <ffff8803f39e9e88>
CR2: 0000000000000004
---[ end trace 5b8e2aa935261552 ]---

I'm still trying to figure out which configuration settings is the cause for the exception. It worked before using 3.7.10 FWIW. zfs and zfs-kmod from master branch.

[ryao]

It looks like autotools did something wrong. Would you paste the config.log from spl?

[posativ]

config.log, spl-9999 uses the latest master from GitHub.

[posativ]

It seems that my autoconf version has changed from 2.13 (working state) to 2.69 (no longer working). Is this an ZFS fault or rather a gentoo build issue?

[posativ]

Moving back to an early "snapshot" of spl (spl-0.6.0_rc14-r1), modprobe zfs solved this issue for me.

[acertain]

NixOS/nixpkgs#39225 might be similar to this? dmesg there, https://gist.github.com/fread2281/c2de6722605838c5d517630274b74b8b is a build log for spl (a different version than the one from the dmesg), and https://gist.github.com/fread2281/c5fd6e556ad02c91b41b198b27acbcce is a build log for zfs

[clefru]

FYI, it seems that current->fs can be NULL under 4.16, see linked NixOS bug.