commit 2865baf54077 ("x86: support user address masking instead of
non-speculative conditional") provided an optimization for
unsafe_get/put_user(), which optimizes the Spectre-V1 mitigation in an
architecture specific way. Currently only x86_64 supports that.
The required code pattern screams for helper functions before it is copied
all over the kernel. So far the exposure is limited to futex, x86 and
fs/select.
Provide a set of helpers for common single size access patterns:
- get/put_user_masked_uNN() where $NN is the variable size in
bits, i.e. 8, 16, 32, 64.
- user_read/write_masked_begin() for scenarios, where several
unsafe_put/get_user() invocations are required.
Convert the existing users over to this.
The series applies on top of
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git locking/urgent
which has a bug fix pending for the futex implementation of this.
It's also available from git:
git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git uaccess/masked
there is another candidate for conversion: RSEQ. This has not been
converted yet because there are more severe performance issues with this
code, which starts to become prominent in exit path profiling because glibc
started to use it. This will be addressed in a seperate series.
As this creates dependencies, merging it should go through one central tree,
i.e. tip, but that's a debate for later :)
Thanks,
tglx
---
arch/x86/include/asm/futex.h | 12 ++----
fs/select.c | 4 --
include/linux/uaccess.h | 78 +++++++++++++++++++++++++++++++++++++++++++
kernel/futex/futex.h | 37 +-------------------
4 files changed, 85 insertions(+), 46 deletions(-)
