linux-kernel - Re: [PATCHv4 0/8] Support for Tegra 2D hardware

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, 03 Jan 2013 10:36:10 +0800
From:	Mark Zhang <nvmarkzhang@...il.com>
To:	Terje Bergström <tbergstrom@...dia.com>
CC:	"thierry.reding@...onic-design.de" <thierry.reding@...onic-design.de>,
	"airlied@...ux.ie" <airlied@...ux.ie>,
	"dev@...xeye.de" <dev@...xeye.de>,
	"dri-devel@...ts.freedesktop.org" <dri-devel@...ts.freedesktop.org>,
	"linux-tegra@...r.kernel.org" <linux-tegra@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCHv4 0/8] Support for Tegra 2D hardware

On 01/02/2013 05:25 PM, Terje Bergström wrote:
> On 26.12.2012 11:42, Mark Zhang wrote:
[...]
> 
>>
>>         if (!de)
>> diff --git a/drivers/gpu/host1x/dev.c b/drivers/gpu/host1x/dev.c
>> index 07e8813..01ed10d 100644
>> --- a/drivers/gpu/host1x/dev.c
>> +++ b/drivers/gpu/host1x/dev.c
>> @@ -38,6 +38,7 @@
>>
>>  struct host1x *host1x;
>>
>> +/* Called by drm unlocked ioctl function. So do we need a lock here? */
>>  void host1x_syncpt_incr_byid(u32 id)
> 
> No, host1x_syncpt_incr_byid is SMP safe.

Correct. Lock is unnecessary.

> 
>>  {
>>         struct host1x_syncpt *sp = host1x->syncpt + id;
>> @@ -52,6 +53,7 @@ u32 host1x_syncpt_read_byid(u32 id)
>>  }
>>  EXPORT_SYMBOL(host1x_syncpt_read_byid);
>>
>> +/* Called by drm unlocked ioctl function. So do we need a lock here? */
>>  int host1x_syncpt_wait_byid(u32 id, u32 thresh, long timeout, u32 *value)
> 
> Same here, SMP safe.
> 
>>  {
>>         struct host1x_syncpt *sp = host1x->syncpt + id;
>> @@ -161,6 +163,8 @@ static int host1x_probe(struct platform_device *dev)
>>
>>         err = host1x_alloc_resources(host);
>>         if (err) {
>> +               /* We don't have chip_ops right now, so here the
>> +                  error message is somewhat improper */
>>                 dev_err(&dev->dev, "failed to init chip support\n");
>>                 goto fail;
>>         }
> 
> Actually, alloc_resources only allocates intr->syncpt, so I the code to
> host1x_intr_init().
> 
>> @@ -175,6 +179,14 @@ static int host1x_probe(struct platform_device *dev)
>>         if (!host->syncpt)
>>                 goto fail;
>>
>> +       /* I suggest create a dedicate function for initializing nop sp.
>> +          First this "_host1x_syncpt_alloc" looks like an internal/static
>> +          function.
>> +          Then actually "host1x_syncpt_alloc" & "_host1x_syncpt_alloc" all
>> +          serve host1x client(e.g: gr2d) so it's not suitable to use them
>> +          for nop sp.
>> +          Just create a wrapper function to call _host1x_syncpt_alloc is OK.
>> +          This will make the code more readable. */
>>         host->nop_sp = _host1x_syncpt_alloc(host, NULL, 0);
> 
> _host1x_syncpt_alloc is an internal function, not exported.
> host1x_syncpt_alloc is exported. I think it's even better if I just move
> allocation of nop_sp to happen in host1x_syncpt_init.
> 

Agree.

>>         if (!host->nop_sp)
>>                 goto fail;
[...]
> 
>> diff --git a/drivers/gpu/host1x/intr.c b/drivers/gpu/host1x/intr.c
>> index efcb9be..e112001 100644
>> --- a/drivers/gpu/host1x/intr.c
>> +++ b/drivers/gpu/host1x/intr.c
>> @@ -329,8 +329,13 @@ void host1x_intr_deinit(struct host1x_intr *intr)
>>  void host1x_intr_start(struct host1x_intr *intr, u32 hz)
>>  {
>>         struct host1x *host1x = intr_to_host1x(intr);
>> +
>> +       /* Why we need to lock here? Seems like this function is
>> +          called by host1x's probe only. */
>>         mutex_lock(&intr->mutex);
>>
>> +       /* "init_host_sync" has already been called in function
>> +          host1x_intr_init. Remove this line. */
>>         host1x->intr_op.init_host_sync(intr);
>>         host1x->intr_op.set_host_clocks_per_usec(intr,
>>                         DIV_ROUND_UP(hz, 1000000));
> 
> In future, we'll call host1x_intr_start() whenever host1x is turned on.
> Thus we need locking.
>
> init_host_sync() should actually be called from host1x_intr_start(), not
> _init().
> 

Got it. Thanks for explanation.

>> diff --git a/drivers/gpu/host1x/syncpt.c b/drivers/gpu/host1x/syncpt.c
>> index 07cbca5..9a234ad 100644
>> --- a/drivers/gpu/host1x/syncpt.c
>> +++ b/drivers/gpu/host1x/syncpt.c
>> @@ -309,6 +309,8 @@ struct host1x_syncpt *host1x_syncpt_init(struct
>> host1x *host)
>>         struct host1x_syncpt *syncpt, *sp;
>>         int i;
>>
>> +       /* Consider devm_kzalloc here. Then you can forget the release
>> +          stuffs about this "syncpt". */
>>         syncpt = sp = kzalloc(sizeof(struct host1x_syncpt) * host->info.nb_pts,
>>                 GFP_KERNEL);
>>         if (!syncpt)
> 
> Will do.
> 
> Thanks!
> 
> Terje
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/