Re: [PATCH v5 3/3] s390/crypto: New s390 specific protected key hash phmac

[Harald Freudenberger <freude@linux.ibm.com>]

On 2024-11-25 07:09, Herbert Xu wrote:
> On Fri, Nov 22, 2024 at 03:34:27PM +0100, Harald Freudenberger wrote:
>> 
>> +static inline int phmac_keyblob2pkey(const u8 *key, unsigned int 
>> keylen,
>> +				     struct phmac_protkey *pk)
>> +{
>> +	int i, rc = -EIO;
>> +
>> +	/* try three times in case of busy card */
>> +	for (i = 0; rc && i < 3; i++) {
>> +		if (rc == -EBUSY && msleep_interruptible(1000))
>> +			return -EINTR;
> 
> You can't sleep in crypto calls in general.
> 
> Now there are specific circumstances where sleeping is allowed,
> e.g., when the MAY_SLEEP flag is set, but I don't think this will
> be acceptable for your purpose.
> 
> I see the same problem exists in paes_s390, which was not reviewed
> on linux-crypto.
> 
> So what exactly causes the EBUSY? The 3 retries followed by failure
> seem rather dangerous too given that this could be used by the block
> layer for critical data that cannot fail.
> 
> Cheers,

Well, I fear the paes implementation is similar broken. The EBUSY
is a reaction of the zcrpyt/AP bus when no eligible crypto card has
been found and a ansynchronous bus scan is running. In that case the
caller should retry in the hope that the bus scan detects a eligible 
card.
Please note that the sleep only happens in in_task() context for the
paes implementation.
For the phmac implementation - didn't you suggest to wrap this with
an asynchronous hmac to be able to sleep in the wrapped synchronous 
hmac?