Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200904165709.GA32667@lst.de> <mhng-5249e999-3e82-417d-8d39-dcb4a159bd83@palmerdabbelt-glaptop1>
 <CAAhSdy0HW8Rjyoiu+Ldx31C9zCBdxJZxhDBcXC4sgitfXnPNDg@mail.gmail.com>
 <CAAhSdy3ESD6L_T1rFprDF2jduN8buTz173F6_mYCbTL3s4RG2A@mail.gmail.com> <20200907061853.GA19038@lst.de>
In-Reply-To: <20200907061853.GA19038@lst.de>
From:   Anup Patel <anup@brainfault.org>
Date:   Mon, 7 Sep 2020 15:29:04 +0530
Message-ID: <CAAhSdy2QkjZXzDBfLpu_Wtn8wWJ+pJSnUiyS+e0e4vtrJ2MWoQ@mail.gmail.com>
Subject: Re: [PATCH] RISC-V: Allow drivers to provide custom read_cycles64 for
 M-mode kernel
To:     Christoph Hellwig <hch@lst.de>
Cc:     Palmer Dabbelt <palmerdabbelt@google.com>,
        Anup Patel <Anup.Patel@wdc.com>,
        Paul Walmsley <paul.walmsley@sifive.com>,
        Albert Ou <aou@eecs.berkeley.edu>,
        Atish Patra <Atish.Patra@wdc.com>,
        Alistair Francis <Alistair.Francis@wdc.com>,
        linux-riscv <linux-riscv@lists.infradead.org>,
        "linux-kernel@vger.kernel.org List" <linux-kernel@vger.kernel.org>,
        Linus Torvalds <torvalds@linux-foundation.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, Sep 7, 2020 at 11:48 AM Christoph Hellwig <hch@lst.de> wrote:
>
> On Sat, Sep 05, 2020 at 11:05:48AM +0530, Anup Patel wrote:
> > Your patch will also break if the SOC specific timer has a 32bit
> > free-running counter
> > unlike the 64bit free-running counter found on CLINT.
> >
> > I guess it's better to let the SOC timer driver provide the
> > method/function to read the
> > free-running counter.
>
> Seriously, build the interfaces once you know the consumers.  Don't
> build pie in the sky interfaces just because you can, because that
> is what creates all the problems.
>
> And of coruse at least for IPIs which absolutely are performance
> criticical we need a standard interface (one that doesn't suck as much
> as the SBI detour with the four extra context switches).  But I guess
> I have already given up on RISC-V because the incompetency about things
> like the irq design are just so horrible that it isn't worth bothering
> any more.

Most of us are aware of these issues. The SBI IPI call will eventually
become fallback option once we have standard mechanism for IPI in
RISC-V privilege spec. Until then we try our best to support all existing
RISC-V systems out there. This also means we end-up supporting
existing SOC specific timers and IPI injection mechanisms.

Regards,
Anup