mbox series

[v3,0/3] crypto: Hash scatterwalk fixes

Message ID cover.1741842470.git.herbert@gondor.apana.org.au
Headers show
Series crypto: Hash scatterwalk fixes | expand

Message

Herbert Xu March 13, 2025, 5:14 a.m. UTC 
v3 restores the krb5 patch and adds kmap_local support to sg miter.
Also remove nth_page since it appears to be unnecessary.

This patch series is based on top of:

https://patchwork.kernel.org/project/linux-crypto/patch/20250310172016.153423-1-ebiggers@kernel.org/

This was prompted by using nth_page instead of incrementing struct
page by hand.  However, that turned out to be unnecessary so what
remains are simply two bug fixes.

Herbert Xu (3):
  lib/scatterlist: Add SG_MITER_LOCAL and use it
  crypto: krb5 - Use SG miter instead of doing it by hand
  crypto: hash - Fix test underflow in shash_ahash_digest

 crypto/ahash.c                   | 40 +++++++++++++++++++++-----------
 crypto/krb5/rfc3961_simplified.c | 35 ++++++++++++----------------
 include/linux/scatterlist.h      |  1 +
 lib/scatterlist.c                | 12 ++++++----
 4 files changed, 50 insertions(+), 38 deletions(-)

Comments

Eric Biggers March 16, 2025, 3:32 a.m. UTC | #1 
On Thu, Mar 13, 2025 at 01:14:53PM +0800, Herbert Xu wrote:
>   * Context:
> - *   May sleep if !SG_MITER_ATOMIC.
> + *   May sleep if !SG_MITER_ATOMIC && !SG_MITER_LOCAL.

This is incorrect.  kmap_local_page() does not disable preemption.

- Eric
Herbert Xu March 16, 2025, 4:34 a.m. UTC | #2 
On Sat, Mar 15, 2025 at 08:32:47PM -0700, Eric Biggers wrote:
> On Thu, Mar 13, 2025 at 01:14:53PM +0800, Herbert Xu wrote:
> >   * Context:
> > - *   May sleep if !SG_MITER_ATOMIC.
> > + *   May sleep if !SG_MITER_ATOMIC && !SG_MITER_LOCAL.
> 
> This is incorrect.  kmap_local_page() does not disable preemption.

I thought it was talking about the function itself, i.e., that
kmap may sleep.  But on a second look yes this is ambiguous and
it could be made clearer.

Cheers,