netdev - Re: [PATCH 03/13] dmaengine: up-level reference counting to the module level

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Thu, 4 Dec 2008 11:51:53 -0700
From:	"Dan Williams" <dan.j.williams@...el.com>
To:	"Guennadi Liakhovetski" <g.liakhovetski@....de>
Cc:	linux-kernel@...r.kernel.org, netdev@...r.kernel.org,
	maciej.sosnowski@...el.com, hskinnemoen@...el.com,
	nicolas.ferre@...el.com
Subject: Re: [PATCH 03/13] dmaengine: up-level reference counting to the module level

On Thu, Dec 4, 2008 at 9:56 AM, Guennadi Liakhovetski
<g.liakhovetski@....de> wrote:
>> +
>> +     /* try to grab channels */
>> +     list_for_each_entry_safe(device, _d, &dma_device_list, global_node)
>> +             list_for_each_entry(chan, &device->channels, device_node) {
>> +                     err = dma_chan_get(chan);
>
> Dan, could you please explain this: dma_chan_get() takes a reference on
> the channel _and_ calls .device_alloc_chan_resources() on first invocation
> for a specific channel. I now see three locations in dmaengine.c, where
> dma_chan_get() is called: in dma_request_channel() - logical, but also in
> dmaengine_get() and dma_async_device_register(), and these latter two I
> don't understand. I do not understand why we have to grab references and
> allocate resources for all (public) channels on all controllers in the
> system if someone just called dmaengine_get()?

Consider the case where a subsystem that consumes engines like
async_tx or net_dma loads before engines are available in the system.
They will have taken a reference against dmaengine, but calls to
dma_find_channel will return NULL.  Once a channel driver is loaded
dma_find_channel() can start returning results, and it is at this
point that resources must be allocated and the backing module pinned.
So dmaengine_get() means "I am interested in offloading stuff, if you
see an offload resource grab it and prep it so I can discover it with
dma_find_channel()".

>
>> @@ -420,6 +443,19 @@ int dma_async_device_register(struct dma_device *device)
>>       }
>>
>>       mutex_lock(&dma_list_mutex);
>> +     list_for_each_entry(chan, &device->channels, device_node) {
>> +             /* if clients are already waiting for channels we need to
>> +              * take references on their behalf
>> +              */
>> +             if (dmaengine_ref_count && dma_chan_get(chan) == -ENODEV) {
>> +                     /* note we can only get here for the first
>> +                      * channel as the remaining channels are
>> +                      * guaranteed to get a reference
>> +                      */
>
> This is the second location - where and how are clients waiting for
> channels? In the old implementation clients had notification callbacks,
> which were called as new channels became available. Now clients are gone,
> so, what is meant here?
>

The assumption is that mem-to-mem offload clients poll dma_find_channel().

Regards,
Dan
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html