Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Thu, 5 Sep 2019 23:32:03 -0700
From:   Christoph Hellwig <hch@infradead.org>
To:     Thomas =?iso-8859-1?Q?Hellstr=F6m_=28VMware=29?= 
        <thomas_os@shipmail.org>
Cc:     Christoph Hellwig <hch@infradead.org>,
        linux-kernel@vger.kernel.org,
        "pv-drivers@vmware.com" <pv-drivers@vmware.com>,
        Thomas Hellstrom <thellstrom@vmware.com>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Christian =?iso-8859-1?Q?K=F6nig?= <christian.koenig@amd.com>
Subject: Re: dma_mmap_fault discussion
Message-ID: <20190906063203.GA25415@infradead.org>
References: <20190905103541.4161-1-thomas_os@shipmail.org>
 <20190905103541.4161-2-thomas_os@shipmail.org>
 <608bbec6-448e-f9d5-b29a-1984225eb078@intel.com>
 <b84d1dca-4542-a491-e585-a96c9d178466@shipmail.org>
 <20190905152438.GA18286@infradead.org>
 <cbbb0e95-8df1-9ab8-59ad-81bd7f3933fa@shipmail.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <cbbb0e95-8df1-9ab8-59ad-81bd7f3933fa@shipmail.org>
User-Agent: Mutt/1.11.4 (2019-03-13)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, Sep 05, 2019 at 07:05:43PM +0200, Thomas Hellstr?m (VMware) wrote:
> I took a quick look at the interfaces and have two questions:
> 
> 1) dma_mmap_prepare(), would it be possible to drop the references to the
> actual coherent region? The thing is that TTM doesn't know at mmap time what
> memory will be backing the region. It can be VRAM, coherent memory or system

I guess we can shift the argument checking into the fault handler.

> 2) @cpu_addr and @dma_addr are the values pointing at the beginning of an
> allocated chunk, right?

Yes.

> The reason I'm asking is that TTM's coherent memory
> pool is sub-allocating from larger chunks into smaller PAGE_SIZE chunks,
> which means that a TTM buffer object may be randomly split across larger
> chunks, which means we have to store these values for each PAGE_SiZE chunk.

For implementations that remap non-contigous regions using vmap we need the
start cpu address passed, as that is used to find the vm_struct structure
for it.  That being said I don't see a problem with you keeping track
of the original start and offset into in your suballocation helpers.