[LTP] ❌ FAIL: Test report for kernel 5.4.0-rc2-d6c2c23.cki (stable-next)

Jan Stancek jstancek@redhat.com
Wed Oct 16 16:38:18 CEST 2019


----- Original Message -----
> 
> From 517d979e84191ae9997c9513a88a5b798af6912f Mon Sep 17 00:00:00 2001
> From: Will Deacon <will@kernel.org>
> Date: Tue, 15 Oct 2019 21:04:18 -0700
> Subject: [PATCH] arm64: tags: Preserve tags for addresses translated via
> TTBR1
> 
> Sign-extending TTBR1 addresses when converting to an untagged address
> breaks the documented POSIX semantics for mlock() in some obscure error
> cases where we end up returning -EINVAL instead of -ENOMEM as a direct
> result of rewriting the upper address bits.
> 
> Rework the untagged_addr() macro to preserve the upper address bits for
> TTBR1 addresses and only clear the tag bits for user addresses. This
> matches the behaviour of the 'clear_address_tag' assembly macro, so
> rename that and align the implementations at the same time so that they
> use the same instruction sequences for the tag manipulation.
> 
> Cc: Catalin Marinas <catalin.marinas@arm.com>
> Link:
> https://lore.kernel.org/stable/20191014162651.GF19200@arrakis.emea.arm.com/
> Reported-by: Jan Stancek <jstancek@redhat.com>
> Signed-off-by: Will Deacon <will@kernel.org>

No regressions observed with LTP syscalls/sched/mm/commands and open_posix_testsuite.

Tested-by: Jan Stancek <jstancek@redhat.com>


More information about the ltp mailing list