Re: [PATCH 2/2] kexec: select CRYPTO from KEXEC_FILE instead of depending on it

["Arnd Bergmann" <arnd@arndb.de>]

On Tue, Oct 24, 2023, at 06:10, Baoquan He wrote:
> On 10/24/23 at 11:55am, Baoquan He wrote:
>> On 10/23/23 at 01:01pm, Arnd Bergmann wrote:
>> > From: Arnd Bergmann <arnd@arndb.de>
>> > 
>> > All other users of crypto code use 'select' instead of 'depends on',
>> > so do the same thing with KEXEC_FILE for consistency.
>> > 
>> > In practice this makes very little difference as kernels with kexec
>> > support are very likely to also include some other feature that already
>> > selects both crypto and crypto_sha256, but being consistent here helps
>> > for usability as well as to avoid potential circular dependencies.
>> > 
>> > This reverts the dependency back to what it was originally before commit
>> > 74ca317c26a3f ("kexec: create a new config option CONFIG_KEXEC_FILE for
>> > new syscall"), which changed changed it with the comment "This should
>>                        ~~~~~~~~~~~~~~ typo
>> > be safer as "select" is not recursive", but that appears to have been
>> > done in error, as "select" is indeed recursive, and there are no other
>> > dependencies that prevent CRYPTO_SHA256 from being selected here.
>> > 
>> > Fixes: 74ca317c26a3f ("kexec: create a new config option CONFIG_KEXEC_FILE for new syscall")
>> > Cc: Herbert Xu <herbert@gondor.apana.org.au>
>> > Cc: "David S. Miller" <davem@davemloft.net>
>> > Cc: linux-crypto@vger.kernel.org
>> > Signed-off-by: Arnd Bergmann <arnd@arndb.de>
>> > ---
>> >  kernel/Kconfig.kexec | 3 ++-
>> >  1 file changed, 2 insertions(+), 1 deletion(-)
>> 
>> LGTM,
>> 
>> Acked-by: Baoquan He <bhe@redhat.com>
>
> Sorry, the patch 1/2 is not sent to me and kexec mailing list, so I
> didn't get the intention of the entire patchset. I need hold the ack
> until I read the patch 1. I have some concerns about patch 1 if I didn't
> misunderstand it. Will come back later when patch 1 reviewing is
> finished.

Sorry about missing you on Cc. If anyone else is looking for the
patch, it's archived at
https://lore.kernel.org/lkml/20231023110308.1202042-1-arnd@kernel.org/

The idea of patch 1 was only to address the build regression on
x86, so I was hoping that part would be uncontroversial. I split
out patch 2/2 since that is intended to actually change the behavior,
hopefully for the better.

I introduced a new regression on riscv that Conor Dooley found, and
that should already be fixed now. It looks like we may need a similar
change on s390

--- a/arch/s390/Kbuild
+++ b/arch/s390/Kbuild
@@ -7,7 +7,7 @@ obj-$(CONFIG_S390_HYPFS)        += hypfs/
 obj-$(CONFIG_APPLDATA_BASE)    += appldata/
 obj-y                          += net/
 obj-$(CONFIG_PCI)              += pci/
-obj-$(CONFIG_ARCH_SUPPORTS_KEXEC_PURGATORY) += purgatory/
+obj-$(CONFIG_KEXEC_FILE)       += purgatory/
 
 # for cleaning
 subdir- += boot tools

but I haven't tested that, and I'll wait for your reply then.

    Arnd