Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 13 Jan 2023 22:29:09 +0100
From:   Joerg Roedel <joro@8bytes.org>
To:     Jason Gunthorpe <jgg@ziepe.ca>
Cc:     Lu Baolu <baolu.lu@linux.intel.com>, Will Deacon <will@kernel.org>,
        Robin Murphy <robin.murphy@arm.com>,
        Vasant Hegde <vasant.hegde@amd.com>,
        Chunyan Zhang <zhang.lyra@gmail.com>,
        Sven Peter <sven@svenpeter.dev>, iommu@lists.linux.dev,
        linux-kernel@vger.kernel.org, Joerg Roedel <jroedel@suse.de>
Subject: Re: [PATCH] iommu/ipmmu-vmsa: Remove ipmmu_utlb_disable()
Message-ID: <Y8HNJfz9R1IidJPq@8bytes.org>
References: <20230113185640.8050-1-joro@8bytes.org>
 <Y8GtFcNq2jugBpI3@ziepe.ca>
 <Y8GwHQLKnXyqa3Ty@8bytes.org>
 <Y8G06onFowE3ZIy8@ziepe.ca>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Y8G06onFowE3ZIy8@ziepe.ca>
Precedence: bulk

On Fri, Jan 13, 2023 at 03:45:46PM -0400, Jason Gunthorpe wrote:
> On Fri, Jan 13, 2023 at 08:25:17PM +0100, Joerg Roedel wrote:
> > In general it is a good idea to at least compile-test every file that is
> > changed in a patch-set before sending it out and not rely on 0-day bot
> > for that.
> 
> Against every arch combination? This is why we have automation bots :(

No, not every combination. But if possible please compile-test each
changed file with a .config that pulls that source file in.  Lots of
drivers can be enabled just with COMPILE_TEST on x86 or be catched with
a generic ARM/ARM64 config which enables all IOMMU drivers.  PAMU is a
bit more difficult as it requires a PPC-32 bit config, but that is the
exception.

A full kernel build is usually also not necessary, often a 'make
drivers/iommu/' with a given config is enough.

That is also how I compile-test the IOMMU tree before I push changes
out. There are per-arch configurations which select all IOMMU drivers on
that arch. Only for X86 I do the full allnoconfig, defconfig,
allmodconfig and allyesconfig cycle (each for 32 and 64 bit).

That certainly does not catch everything, but a lot of compile issues can be
found that way. And for patch-sets only touching, for example, VT-d it
is still enough to only compile-test on x86. A patch-set touching that
much drivers is rather the exception.

Regards,

	Joerg