selftests/x86/ldt_gdt: Robustify against set_thread_area() and LAR oddities
authorAndy Lutomirski <luto@kernel.org>
Sat, 4 Nov 2017 11:19:49 +0000 (04:19 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 10 Dec 2017 12:40:39 +0000 (13:40 +0100)
commit898fe968f78c45e12fed31aee35a118a4e26e1b4
treee3d05f1c9d4c7d70e3e1db497a91d47f44221828
parent9aaa793b6b05b57cf37814ccf0d992035ea41407
selftests/x86/ldt_gdt: Robustify against set_thread_area() and LAR oddities

[ Upstream commit d60ad744c9741586010d4bea286f09a063a90fbd ]

Bits 19:16 of LAR's result are undefined, and some upcoming
improvements to the test case seem to trigger this.  Mask off those
bits to avoid spurious failures.

commit 5b781c7e317f ("x86/tls: Forcibly set the accessed bit in TLS
segments") adds a valid case in which LAR's output doesn't quite
agree with set_thread_area()'s input.  This isn't triggered in the
test as is, but it will be if we start calling set_thread_area()
with the accessed bit clear.  Work around this discrepency.

I've added a Fixes tag so that -stable can pick this up if neccesary.

Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bpetkov@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 5b781c7e317f ("x86/tls: Forcibly set the accessed bit in TLS segments")
Link: http://lkml.kernel.org/r/b82f3f89c034b53580970ac865139fd8863f44e2.1509794321.git.luto@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
tools/testing/selftests/x86/ldt_gdt.c