[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <Ya88fFJC4Kdjyb+B@arm.com>
Date: Tue, 7 Dec 2021 10:50:36 +0000
From: Catalin Marinas <catalin.marinas@....com>
To: Yunfeng Ye <yeyunfeng@...wei.com>
Cc: will@...nel.org, wangkefeng.wang@...wei.com,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
wuxu.wu@...wei.com, Hewenliang <hewenliang4@...wei.com>
Subject: Re: [PATCH] arm64: mm: Use asid2idx() and asid feature macro for
cleanup
On Tue, Dec 07, 2021 at 10:21:26AM +0800, Yunfeng Ye wrote:
> On 2021/12/7 0:23, Catalin Marinas wrote:
> > On Thu, Oct 28, 2021 at 08:27:23PM +0800, Yunfeng Ye wrote:
> >> @@ -162,7 +162,7 @@ static u64 new_context(struct mm_struct *mm)
> >> u64 generation = atomic64_read(&asid_generation);
> >>
> >> if (asid != 0) {
> >> - u64 newasid = generation | (asid & ~ASID_MASK);
> >> + u64 newasid = generation | asid2idx(asid);
> >>
> >> /*
> >> * If our current ASID was active during a rollover, we
> >> @@ -306,7 +306,7 @@ unsigned long arm64_mm_context_get(struct mm_struct *mm)
> >> out_unlock:
> >> raw_spin_unlock_irqrestore(&cpu_asid_lock, flags);
> >>
> >> - asid &= ~ASID_MASK;
> >> + asid = asid2idx(asid);
> >
> > While functionally the code is the same, I don't think this was the
> > intention of asid2idx(). It's meant to provide an index into asid_map,
> > while the ASID_MASK lines isolate the asid number and add a new
> > generation to it.
>
> The commit 0c8ea531b774 ("arm64: mm: Allocate ASIDs in pairs") introduce the
> asid2idx and idx2asid macro, but these macros is not really useful after the
> commit f88f42f853a8 ("arm64: context: Free up kernel ASIDs if KPTI is not in use").
>
> I think "(asid & ~ASID_MASK)" can be instead by a macro, it is the same code with
> asid2idx(). Can it be renamed? (eg, ctxid2asid)
Yes, that would work.
--
Catalin
Powered by blists - more mailing lists