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: <20210713104026.58560-1-andriy.shevchenko@linux.intel.com>
 <20210713104026.58560-3-andriy.shevchenko@linux.intel.com> <9af24b96-8119-7ccf-f0d0-d725af80aa0b@kernel.org>
In-Reply-To: <9af24b96-8119-7ccf-f0d0-d725af80aa0b@kernel.org>
From:   Andy Shevchenko <andy.shevchenko@gmail.com>
Date:   Wed, 14 Jul 2021 12:15:25 +0300
Message-ID: <CAHp75VeuG08M9nURpEmW79euKSJkYvLnFiUe+6cGpRHL4zUOfw@mail.gmail.com>
Subject: Re: [PATCH v1 3/4] serial: 8250_pci: Always try MSI/MSI-X
To:     Jiri Slaby <jirislaby@kernel.org>, Randy Wright <rwright@hpe.com>
Cc:     Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        linux-serial@vger.kernel.org, linux-kernel@vger.kernel.org,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Ralf Ramsauer <ralf.ramsauer@oth-regensburg.de>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Jul 14, 2021 at 9:55 AM Jiri Slaby <jirislaby@kernel.org> wrote:
>
> On 13. 07. 21, 12:40, Andy Shevchenko wrote:
> > There is no need to try MSI/MSI-X only on selected devices.
> > If MSI is not supported while neing advertised it means device
>
> being
>
> > is broken and we rather introduce a list of such devices which
> > hopefully will be small or never appear.
>
> Hmm, have you checked the commit which introduced the whitelist?

Nope, my bad.

>      Nevertheless, this needs to handled with care: while many 8250 devices
>      actually claim to support MSI(-X) interrupts it should not be
> enabled be
>      default. I had at least one device in my hands with broken MSI
>      implementation.
>
>      So better introduce a whitelist with devices that are known to support
>      MSI(-X) interrupts. I tested all devices mentioned in the patch.

Thanks, but I still think that blacklisting is better. All drivers I
have split (or participated in splitting) from 8250_pci have enabled
MSI for the entire subset they serve for.

> You should have at least CCed the author for an input.

Thanks. I also added Randy, who extended the list.

...

> > +             pci_set_master(dev);
>
> But bus mastering is not about MSIs.

Strictly speaking it's not, but MSI won't work without DMA.

>  I *think* it's still OK, but you
> need to document that in the commit log too.
>
> Actually, why the commit which added this code turns on bus mastering?

-- 
With Best Regards,
Andy Shevchenko