linux-kernel - Re: [PATCH] riscv: Enable interrupt during exception handling

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: <20250624141801.GV1613200@noisy.programming.kicks-ass.net>
Date: Tue, 24 Jun 2025 16:18:01 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Cc: Clément Léger <cleger@...osinc.com>,
	Nam Cao <namcao@...utronix.de>,
	"Maciej W. Rozycki" <macro@...am.me.uk>,
	Palmer Dabbelt <palmer@...belt.com>,
	Paul Walmsley <paul.walmsley@...ive.com>, aou@...s.berkeley.edu,
	Alexandre Ghiti <alex@...ti.fr>, clrkwllms@...nel.org,
	rostedt@...dmis.org, linux-rt-devel@...ts.linux.dev,
	linux-riscv@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] riscv: Enable interrupt during exception handling

On Tue, Jun 24, 2025 at 04:11:30PM +0200, Sebastian Andrzej Siewior wrote:
> On 2025-06-24 16:08:15 [+0200], Peter Zijlstra wrote:
> > > >>
> > > >> I also had a series which was doing so for misaligned accesses handling,
> > > >> but after discussion, it was not ok to do so.:
> > > >>
> > > >> https://lore.kernel.org/linux-riscv/20250422094419.GC14170@noisy.programming.kicks-ass.net/
> > > > 
> > > > If I understand that right, exceptions from kernel should be treated as
> > > > NMI, so that lockdep can tell us if exception handlers touch locks.
> > > > 
> > > > But (conditionally) enabling interrupts does not lose us that benefit. It
> > > > is still considered NMI by lockdep.
> > > > 
> > > > Unless I miss something, the patch is fine as is.
> > 
> > I'm confused, you're wanting to conditionally enable interrupts from a
> > kernel exception while its NMI like? *WHY* ?!
> 
> What we want is to enable interrupt handling if it was enabled before
> the exception occured. So we can send a proper signal on PREEMPT_RT
> without chocking on spinlock_t/ sighand.

I'm confused, sending signals is for exception from userspace. That has
nothing to do with exceptions from kernelspace being NMI like.