linux-kernel - Re: [PATCH v5 2/8] firmware: add support for ARM System Control and Power Interface(SCPI) protocol

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]

Message-ID: <CABb+yY0zy4w02Q3B63HfdT8YyOH_f_SGrSZ-8r1tGPHCf2d_9A@mail.gmail.com>
Date:	Fri, 31 Jul 2015 19:15:01 +0530
From:	Jassi Brar <jassisinghbrar@...il.com>
To:	Sudeep Holla <sudeep.holla@....com>
Cc:	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Liviu Dudau <Liviu.Dudau@....com>,
	Punit Agrawal <Punit.Agrawal@....com>,
	"Jon Medhurst (Tixy)" <tixy@...aro.org>,
	Lorenzo Pieralisi <Lorenzo.Pieralisi@....com>,
	Arnd Bergmann <arnd@...db.de>, Olof Johansson <olof@...om.net>,
	Kevin Hilman <khilman@...nel.org>
Subject: Re: [PATCH v5 2/8] firmware: add support for ARM System Control and
 Power Interface(SCPI) protocol

On Fri, Jul 31, 2015 at 6:38 PM, Sudeep Holla <sudeep.holla@....com> wrote:
>
>
> On 31/07/15 11:43, Sudeep Holla wrote:
>>
>>
>>
>> On 31/07/15 11:38, Jassi Brar wrote:
>>>
>>> On Fri, Jul 31, 2015 at 3:10 PM, Sudeep Holla <sudeep.holla@....com>
>>> wrote:
>>>>
>>>>
>>>> I forgot to mention, we have a the following description in
>>>> mbox_client_txdone which is misleading:
>>>>
>>>> "The client/protocol had received some 'ACK' packet and it notifies the
>>>> API that the last packet was sent successfully. This only works if the
>>>> controller can't sense TX-Done."
>>>>
>>>> which is clearly not the case in SCPI. IMO we may have to reword that.
>>>>
>>> Yes. And also see whether it could race with polling driven tx_tick.
>>>
>>
>> Yes I am also looking at that now while I am trying to check if
>> TXDONE_BY_ACK works on Juno, will keep you posted.
>>
>
> OK, I recollect the racy condition now which I had in my mind from the
> beginning convincing myself why we can't use that option. I was not good
> in words to explain it so far but let me try with the ASCII art this
> time. Note Tx ACK below means the remote setting the register flag and
> not to be confused with the ACK packet. For simplicity Rx can be assumed
> to be Tx ACK packet
>
>      Time       MHU/SCPI                        Remote SCP
>                    |                                |
>         1          |------------ Tx1 -------------->|
>                    |                                |
>         2          |<----------- Tx1 ACK -----------|
>                    |                                |
>         3          |------------ Tx2 -------------->|
>                    |                                |
>         4          |<----------- Rx1 ---------------|
>                    |                                |
>         5          |<----------- Tx2 ACK -----------|
>                    |                                |
>         6          |------------ Tx3 -------------->|
>                    |                                |
>         7          |<----------- Rx2 ---------------|
>
> Now lets consider the above scenario, suppose we have TXDONE_BY_ACK
> and use mbox_client_txdone in Rx interrupt(i.e. response packet), we end
> up in the race easily IIUC.
>
> E.g. A client would have sent Tx2(3) before Rx1 interrupt(4) and if we
> ACK Tx1 now in (3), we will end up acknowledging Tx2(5) as the mailbox
> core assumes only one Tx request at a time which is clearly not the case
> in our setup. The client can then go ahead and send Tx3(6) overwriting
> the payload while remote was processing or even result in remote missing
> the request completely. Does that make sense or am I missing something ?
>
Yeah, that could happen. But the race can be fixed by checking
last_tx_done if the controller provides that. If last_tx_done is not
implemented, polling won't race.

Please try the following...

==================>8=================
diff --git a/drivers/mailbox/mailbox.c b/drivers/mailbox/mailbox.c
index 07fd507..c973c2c 100644
--- a/drivers/mailbox/mailbox.c
+++ b/drivers/mailbox/mailbox.c
@@ -181,17 +181,23 @@ EXPORT_SYMBOL_GPL(mbox_chan_txdone);
  * @r: Success status of last transmission.
  *
  * The client/protocol had received some 'ACK' packet and it notifies
- * the API that the last packet was sent successfully. This only works
- * if the controller can't sense TX-Done.
+ * the API that the last packet was sent successfully.
  */
 void mbox_client_txdone(struct mbox_chan *chan, int r)
 {
+       bool txdone = true;
+
        if (unlikely(!(chan->txdone_method & TXDONE_BY_ACK))) {
                dev_err(chan->mbox->dev, "Client can't run the TX ticker\n");
                return;
        }

-       tx_tick(chan, r);
+       /* if h/w can check for tx-done, make sure its done */
+       if (chan->mbox->ops->last_tx_done)
+               txdone = chan->mbox->ops->last_tx_done(chan);
+
+       if (txdone)
+               tx_tick(chan, 0);
 }
 EXPORT_SYMBOL_GPL(mbox_client_txdone);

==================8<=================

we might also have to looking into protecting last_tx_done and tx_tick
together. Anyways, those changes are warranted in order to get the
expected functionality.

Thanks.
--
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/