linux-kernel - RE: [PATCH] x86/mce: Dynamically size space for machine check records

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

Message-ID: <SJ1PR11MB60831369FC6FEA7944DBEEC5FC5F2@SJ1PR11MB6083.namprd11.prod.outlook.com>
Date: Thu, 29 Feb 2024 18:38:58 +0000
From: "Luck, Tony" <tony.luck@...el.com>
To: "Naik, Avadhut" <avadnaik@....com>, Borislav Petkov <bp@...en8.de>
CC: "Mehta, Sohil" <sohil.mehta@...el.com>, "x86@...nel.org" <x86@...nel.org>,
	"linux-edac@...r.kernel.org" <linux-edac@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"yazen.ghannam@....com" <yazen.ghannam@....com>, Avadhut Naik
	<avadhut.naik@....com>
Subject: RE: [PATCH] x86/mce: Dynamically size space for machine check records

> Wouldn't having dedup actually increase the time we spend #MC context?
> Comparing the new MCE record against each existing record in the
> genpool.

Yes, dedup would take extra time (increasing linearly with the number
of pending errors that were not filtered out by the dedup process).

> AFAIK, MCEs cannot be nested. Correct me if I am wrong here.

Can't be nested on the same CPU. But multiple CPUs may take
a local machine check simultaneously. Local machine check is
opt-in on Intel, I believe it is default on AMD.

Errors can also be signaled with CMCI.

> In a flood situation, like the one described above, that is exactly
> what may happen: An MCE coming in while the dedup mechanism is
> underway (in #MC context).

In a flood of errors it would be complicated to synchronize dedup filtering
on multiple CPUs. The trade-off between trying to get that code right,
and just allocating a few extra Kbytes of memory would seem to favor
allocating more memory.

-- 
Thanks,
Avadhut Naik