linux-kernel - x86_64: can boxen lose IPIs during IPI storm?

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:	Wed, 22 Aug 2012 14:59:16 +0200
From:	Mike Galbraith <efault@....de>
To:	LKML <linux-kernel@...r.kernel.org>
Cc:	Ingo Molnar <mingo@...e.hu>
Subject: x86_64: can boxen lose IPIs during IPI storm?

Greetings,

Not so hypothetical situation: CPUs are having a major SCSI induced IPI
fest, printk via serial port is happening while the storm is raging.
Can it happen that while IRQs are disabled, multiple IPIs arrive, thus
losing one or more?  Seems like that possibility _has_ to be excluded by
hardware for SMP to actually work (and it generally does;) but...

I ask because of a dump I saw, where CPU3 waits for CPU1 forever after
having IPId all CPUs, with all except CPU1 having responded.  CPU1,
seemingly unaware of any such IPI, was then used to trigger that very
dump, ie was not busted/stuck, which looked rather like a lost IPI.

-Mike

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/