Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1662865 > unrolled thread

Re: [PATCH v2 01/11] crypto: introduce crypto wait for async op

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH v2 01/11] crypto: introduce crypto wait for async op Herbert Xu <herbert@gondor.apana.org.au> - 2017-06-10 05:50 +0200
    Re: [PATCH v2 01/11] crypto: introduce crypto wait for async op Gilad Ben-Yossef <gilad@benyossef.com> - 2017-06-10 10:10 +0200
      Re: [PATCH v2 01/11] crypto: introduce crypto wait for async op Herbert Xu <herbert@gondor.apana.org.au> - 2017-06-10 11:10 +0200

#1662865 — Re: [PATCH v2 01/11] crypto: introduce crypto wait for async op

On Mon, May 29, 2017 at 11:22:48AM +0300, Gilad Ben-Yossef wrote:
>
> +static inline int crypto_wait_req(int err, struct crypto_wait *wait)
> +{
> +	switch (err) {
> +	case -EINPROGRESS:
> +	case -EBUSY:
> +		wait_for_completion(&wait->completion);
> +		reinit_completion(&wait->completion);
> +		err = wait->err;
> +		break;
> +	};
> +
> +	return err;
> +}

This assumes that the request is used with backlog.  For non-backlog
requests this would result in a memory leak as EBUSY in that case is
a fatal error.

So this API can't be used without backlog.

We could introduce a flag to indicate whether we want backlog or not,
or maybe we should change our API so that in the non-backlog case we
return something other than EBUSY.

Opinions?

Thanks,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

[toc] | [next] | [standalone]

#1662899

On Sat, Jun 10, 2017 at 6:43 AM, Herbert Xu <herbert@gondor.apana.org.au> wrote:
> On Mon, May 29, 2017 at 11:22:48AM +0300, Gilad Ben-Yossef wrote:
>>
>> +static inline int crypto_wait_req(int err, struct crypto_wait *wait)
>> +{
>> +     switch (err) {
>> +     case -EINPROGRESS:
>> +     case -EBUSY:
>> +             wait_for_completion(&wait->completion);
>> +             reinit_completion(&wait->completion);
>> +             err = wait->err;
>> +             break;
>> +     };
>> +
>> +     return err;
>> +}
>
> This assumes that the request is used with backlog.  For non-backlog
> requests this would result in a memory leak as EBUSY in that case is
> a fatal error.
>
> So this API can't be used without backlog.

You are right, of course. I did not take that into account.

>
> We could introduce a flag to indicate whether we want backlog or not,
> or maybe we should change our API so that in the non-backlog case we
> return something other than EBUSY.
>
> Opinions?

I guess there is a question if it really is important to know that
your request ended up
on the backlog, rather than being handled.I can imagine it can be used
as back pressure
indication but I wonder if someone is using that.

If not, maybe we can simplify things and use EINPROGRESS asindication
of a request
being accepted by the next layer (either being processed or queued in
the back log), whereas
EBUSY would indicate failure.

It does have a potential to make things simpler, I think.

Gilad

>
> Thanks,
> --
> Email: Herbert Xu <herbert@gondor.apana.org.au>
> Home Page: http://gondor.apana.org.au/~herbert/
> PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt



-- 
Gilad Ben-Yossef
Chief Coffee Drinker

"If you take a class in large-scale robotics, can you end up in a
situation where the homework eats your dog?"
 -- Jean-Baptiste Queru

[toc] | [prev] | [next] | [standalone]

#1662914

On Sat, Jun 10, 2017 at 11:05:39AM +0300, Gilad Ben-Yossef wrote:
>
> I guess there is a question if it really is important to know that
> your request ended up
> on the backlog, rather than being handled.I can imagine it can be used
> as back pressure
> indication but I wonder if someone is using that.

Oh yes we do want it to return EBUSY if we put it on the backlog
because in that case we want the user to stop sending us new
requests.

It's the other case where we dropped the request and returned
EBUSY where I think we could return something other than EBUSY
and get rid of the ambiguity.

Thanks,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web