On Wed, Dec 11, 2024 at 05:30:56PM +0800, Herbert Xu wrote:
> On Tue, Dec 10, 2024 at 09:48:33PM +0100, Christian Marangi wrote:
> >
> > + /*
> > + * Consume remaining data.
> > + * 1. Loop until we consume all the data in block of 64bytes
> > + * 2. Send full block of 64bytes
> > + * 3. Skip sending last block for future update() or for final() to
> > + * enable HASH_FINALIZE bit.
> > + */
> > + while (to_consume > 0) {
> > + int to_read = min(to_consume, SHA256_BLOCK_SIZE);
> > +
> > + block = kzalloc(sizeof(*block), GFP_KERNEL);
>
> You should avoid allocating memory. If you really must do it,
> then it needs to be GFP_ATOMIC, and your algorithm needs to set
> CRYPTO_ALG_ALLOCATES_MEMORY which means that it won't be used
> by the storage layer as memory allocations may lead to dead-lock.
>
It's problematic for the ring consumption logic to free the block as
they get consumed hence memory allocation is needed.
> The preferred way to access extra memory is through the request
> context structure.
>
Do you have quick example for this?
--
Ansuel
