linux-kernel - about prandom vs __this_cpu

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Date:   Tue, 13 Oct 2020 04:55:29 +0200
From:   Willy Tarreau <w@....eu>
To:     Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: about prandom vs __this_cpu_read()

Linus,

the latest prandom series I sent you changes __this_cpu_read() to
this_cpu_read() since LKP returned the following trace on i386:

[    9.061912]  dump_stack+0x54/0x6e
[    9.061912]  check_preemption_disabled+0xb1/0xc0
[    9.061912]  __this_cpu_preempt_check+0xf/0x11
[    9.065516]  __igt_reserve+0x91/0x580
[    9.065516]  igt_reserve+0x5b/0xc0
[    9.065516]  test_drm_mm_init+0x78/0xa9

I got another report that it breaks on ARC (trivial fix consists in
adding irqflags.h in prandom.h) but that made me wonder, we're using
this to access and update the net_rand_noise variable only, wouldn't
it be better to use the faster raw_cpu_read()/raw_cpu_write() there,
considering we don't care about concurrent accesses since they simply
add even more impredictability to the noise ?

If so, I'll update this branch and will send you another PR.

Thanks,
Willy