Re: [PATCH RT v2 0/3] riscv: add PREEMPT_RT support

From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To: Jisheng Zhang <jszhang@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>,
	Paul Walmsley <paul.walmsley@sifive.com>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Albert Ou <aou@eecs.berkeley.edu>,
	linux-riscv@lists.infradead.org, tobias.schaffner@siemens.com,
	Arnd Bergmann <arnd@arndb.de>,
	Conor Dooley <conor.dooley@microchip.com>
Subject: Re: [PATCH RT v2 0/3] riscv: add PREEMPT_RT support
Date: Tue, 31 Oct 2023 17:44:11 +0100	[thread overview]
Message-ID: <20231031164411.9YRGasL-@linutronix.de> (raw)
In-Reply-To: <ZUEiCUxcd11inlRP@xhacker>

On 2023-10-31 23:49:29 [+0800], Jisheng Zhang wrote:
> Yes there's no third patch. I didn't use the correct number in patch0's
> subject.

So it looks fine. The warning was CPU-hotplug related
	https://lore.kernel.org/all/0abd0acf-70a1-d546-a517-19efe60042d1@microchip.com/

and it looks to be gone as of commit
	5944ce092b97c ("arch_topology: Build cacheinfo from primary CPU")

so that good. Any double checking is welcome of course ;)
JUMP_LABELs don't use stop_cpu. Check.
The timer is PERCPU. Check.
Can't find perf events. But the commit for threaded interrupts claims to
have them per-CPU. 
Has HAVE_POSIX_CPU_TIMERS_TASK_WORK with generic kvm. Check.

die() and die_lock. It looks like die_lock is acquired when the system
is done and requires medical assistance. This would qualify it for a
raw_spinlock_t. Also, should any of the bad things happen in a section
with disabled preemption or interrupts then a spinlock_t can not be
acquired. Unless die() is always invoked in a preemptible context…

The other things are covered by the generic code. I think I didn't miss
anything…
I going to have new release by the end of the week at the latest with
this bits. Please look after the die_lock.

Sebastian

_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv