For historical reasons, the arm64 kernel VA space is configured as two
equally sized halves, i.e., on a 48-bit VA build, the VA space is split
into a 47-bit vmalloc region and a 47-bit linear region.
When support for 52-bit virtual addressing was added, this equal split
was kept, resulting in a substantial waste of virtual address space in
the linear region:
48-bit VA 52-bit VA
0xffff_ffff_ffff_ffff +-------------+ +-------------+
| vmalloc | | vmalloc |
0xffff_8000_0000_0000 +-------------+ _PAGE_END(48) +-------------+
| linear | : :
0xffff_0000_0000_0000 +-------------+ : :
: : : :
: : : :
: : : :
: : : currently :
: unusable : : :
: : : unused :
: by : : :
: : : :
: hardware : : :
: : : :
0xfff8_0000_0000_0000 : : _PAGE_END(52) +-------------+
: : | |
: : | |
: : | |
: : | |
: : | |
: unusable : | |
: : | linear |
: by : | |
: : | region |
: hardware : | |
: : | |
: : | |
: : | |
: : | |
: : | |
: : | |
0xfff0_0000_0000_0000 +-------------+ PAGE_OFFSET +-------------+
As illustrated above, the 52-bit VA kernel uses 47 bits for the vmalloc
space (as before), to ensure that a single 64k granule kernel image can
support any 64k granule capable system, regardless of whether it supports
the 52-bit virtual addressing extension. However, due to the fact that
the VA space is still split in equal halves, the linear region is only
2^51 bytes in size, wasting almost half of the 52-bit VA space.
Let's fix this, by abandoning the equal split, and simply assigning all
VA space outside of the vmalloc region to the linear region.
The KASAN shadow region is reconfigured so that it ends at the start of
the vmalloc region, and grows downwards. That way, the arrangement of
the vmalloc space (which contains kernel mappings, modules, BPF region,
the vmemmap array etc) is identical between non-KASAN and KASAN builds,
which aids debugging.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Steve Capper <steve.capper@arm.com>
Link: https://lore.kernel.org/r/20201008153602.9467-3-ardb@kernel.org
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
KASAN_SHADOW_OFFSET is a constant that is supplied to gcc as a command
line argument and affects the codegen of the inline address sanetiser.
Essentially, for an example memory access:
*ptr1 = val;
The compiler will insert logic similar to the below:
shadowValue = *(ptr1 >> KASAN_SHADOW_SCALE_SHIFT + KASAN_SHADOW_OFFSET)
if (somethingWrong(shadowValue))
flagAnError();
This code sequence is inserted into many places, thus
KASAN_SHADOW_OFFSET is essentially baked into many places in the kernel
text.
If we want to run a single kernel binary with multiple address spaces,
then we need to do this with KASAN_SHADOW_OFFSET fixed.
Thankfully, due to the way the KASAN_SHADOW_OFFSET is used to provide
shadow addresses we know that the end of the shadow region is constant
w.r.t. VA space size:
KASAN_SHADOW_END = ~0 >> KASAN_SHADOW_SCALE_SHIFT + KASAN_SHADOW_OFFSET
This means that if we increase the size of the VA space, the start of
the KASAN region expands into lower addresses whilst the end of the
KASAN region is fixed.
Currently the arm64 code computes KASAN_SHADOW_OFFSET at build time via
build scripts with the VA size used as a parameter. (There are build
time checks in the C code too to ensure that expected values are being
derived). It is sufficient, and indeed is a simplification, to remove
the build scripts (and build time checks) entirely and instead provide
KASAN_SHADOW_OFFSET values.
This patch removes the logic to compute the KASAN_SHADOW_OFFSET in the
arm64 Makefile, and instead we adopt the approach used by x86 to supply
offset values in kConfig. To help debug/develop future VA space changes,
the Makefile logic has been preserved in a script file in the arm64
Documentation folder.
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Ard Biesheuvel