talos-obmc-linux/arch/arm/lib, branch dev-5.0

ARM: 8833/1: Ensure that NEON code always compiles with Clang

2019-04-05T20:34:39+00:00

[ Upstream commit de9c0d49d85dc563549972edc5589d195cd5e859 ] While building arm32 allyesconfig, I ran into the following errors: arch/arm/lib/xor-neon.c:17:2: error: You should compile this file with '-mfloat-abi=softfp -mfpu=neon' In file included from lib/raid6/neon1.c:27: /home/nathan/cbl/prebuilt/lib/clang/8.0.0/include/arm_neon.h:28:2: error: "NEON support not enabled" Building V=1 showed NEON_FLAGS getting passed along to Clang but __ARM_NEON__ was not getting defined. Ultimately, it boils down to Clang only defining __ARM_NEON__ when targeting armv7, rather than armv6k, which is the '-march' value for allyesconfig. >From lib/Basic/Targets/ARM.cpp in the Clang source: // This only gets set when Neon instructions are actually available, unlike // the VFP define, hence the soft float and arch check. This is subtly // different from gcc, we follow the intent which was that it should be set // when Neon instructions are actually available. if ((FPU & NeonFPU) && !SoftFloat && ArchVersion >= 7) { Builder.defineMacro("__ARM_NEON", "1"); Builder.defineMacro("__ARM_NEON__"); // current AArch32 NEON implementations do not support double-precision // floating-point even when it is present in VFP. Builder.defineMacro("__ARM_NEON_FP", "0x" + Twine::utohexstr(HW_FP & ~HW_FP_DP)); } Ard Biesheuvel recommended explicitly adding '-march=armv7-a' at the beginning of the NEON_FLAGS definitions so that __ARM_NEON__ always gets definined by Clang. This doesn't functionally change anything because that code will only run where NEON is supported, which is implicitly armv7. Link: https://github.com/ClangBuiltLinux/linux/issues/287 Suggested-by: Ard Biesheuvel Signed-off-by: Nathan Chancellor Acked-by: Nicolas Pitre Reviewed-by: Nick Desaulniers Reviewed-by: Stefan Agner Signed-off-by: Russell King Signed-off-by: Sasha Levin

ARM: 8813/1: Make aligned 2-byte getuser()/putuser() atomic on ARMv6+

2018-11-12T10:52:04+00:00

getuser() and putuser() (and there underscored variants) use two strb[t]/ldrb[t] instructions when they are asked to get/put 16-bits. This means that the read/write is not atomic even when performed to a 16-bit-aligned address. This leads to problems with vhost: vhost uses __getuser() to read the vring's 16-bit avail.index field, and if it happens to observe a partial update of the index, wrong descriptors will be used which will lead to a breakdown of the virtio communication. A similar problem exists for __putuser() which is used to write to the vring's used.index field. The reason these functions use strb[t]/ldrb[t] is because strht/ldrht instructions did not exist until ARMv6T2/ARMv7. So we should be easily able to fix this on ARMv7. Also, since all ARMv6 processors also don't actually use the unprivileged instructions anymore for uaccess (since CONFIG_CPU_USE_DOMAINS is not used) we can easily fix them too. Signed-off-by: Vincent Whitchurch Signed-off-by: Russell King

ARM: 8812/1: Optimise copy_{from/to}_user for !CPU_USE_DOMAINS

2018-11-12T10:51:59+00:00

ARMv6+ processors do not use CONFIG_CPU_USE_DOMAINS and use privileged ldr/str instructions in copy_{from/to}_user. They are currently unnecessarily using single ldr/str instructions and can use ldm/stm instructions instead like memcpy does (but with appropriate fixup tables). This speeds up a "dd if=foo of=bar bs=32k" on a tmpfs filesystem by about 4% on my Cortex-A9. before:134217728 bytes (128.0MB) copied, 0.543848 seconds, 235.4MB/s before:134217728 bytes (128.0MB) copied, 0.538610 seconds, 237.6MB/s before:134217728 bytes (128.0MB) copied, 0.544356 seconds, 235.1MB/s before:134217728 bytes (128.0MB) copied, 0.544364 seconds, 235.1MB/s before:134217728 bytes (128.0MB) copied, 0.537130 seconds, 238.3MB/s before:134217728 bytes (128.0MB) copied, 0.533443 seconds, 240.0MB/s before:134217728 bytes (128.0MB) copied, 0.545691 seconds, 234.6MB/s before:134217728 bytes (128.0MB) copied, 0.534695 seconds, 239.4MB/s before:134217728 bytes (128.0MB) copied, 0.540561 seconds, 236.8MB/s before:134217728 bytes (128.0MB) copied, 0.541025 seconds, 236.6MB/s after:134217728 bytes (128.0MB) copied, 0.520445 seconds, 245.9MB/s after:134217728 bytes (128.0MB) copied, 0.527846 seconds, 242.5MB/s after:134217728 bytes (128.0MB) copied, 0.519510 seconds, 246.4MB/s after:134217728 bytes (128.0MB) copied, 0.527231 seconds, 242.8MB/s after:134217728 bytes (128.0MB) copied, 0.525030 seconds, 243.8MB/s after:134217728 bytes (128.0MB) copied, 0.524236 seconds, 244.2MB/s after:134217728 bytes (128.0MB) copied, 0.523659 seconds, 244.4MB/s after:134217728 bytes (128.0MB) copied, 0.525018 seconds, 243.8MB/s after:134217728 bytes (128.0MB) copied, 0.519249 seconds, 246.5MB/s after:134217728 bytes (128.0MB) copied, 0.518527 seconds, 246.9MB/s Reviewed-by: Nicolas Pitre Signed-off-by: Vincent Whitchurch Signed-off-by: Russell King

Merge branches 'fixes', 'misc' and 'spectre' into for-next

2018-10-10T12:53:33+00:00

ARM: 8797/1: spectre-v1.1: harden __copy_to_user

2018-10-05T09:51:15+00:00

Sanitize user pointer given to __copy_to_user, both for standard version and memcopy version of the user accessor. Signed-off-by: Julien Thierry Signed-off-by: Russell King

ARM: 8796/1: spectre-v1,v1.1: provide helpers for address sanitization

2018-10-05T09:51:15+00:00

Introduce C and asm helpers to sanitize user address, taking the address range they target into account. Use asm helper for existing sanitization in __copy_from_user(). Signed-off-by: Julien Thierry Signed-off-by: Russell King

Merge branches 'fixes', 'misc' and 'spectre' into for-linus

2018-08-13T15:28:50+00:00

Conflicts: arch/arm/include/asm/uaccess.h Signed-off-by: Russell King

ARM: spectre-v1: mitigate user accesses

2018-08-02T16:41:38+00:00

Spectre variant 1 attacks are about this sequence of pseudo-code: index = load(user-manipulated pointer); access(base + index * stride); In order for the cache side-channel to work, the access() must me made to memory which userspace can detect whether cache lines have been loaded. On 32-bit ARM, this must be either user accessible memory, or a kernel mapping of that same user accessible memory. The problem occurs when the load() speculatively loads privileged data, and the subsequent access() is made to user accessible memory. Any load() which makes use of a user-maniplated pointer is a potential problem if the data it has loaded is used in a subsequent access. This also applies for the access() if the data loaded by that access is used by a subsequent access. Harden the get_user() accessors against Spectre attacks by forcing out of bounds addresses to a NULL pointer. This prevents get_user() being used as the load() step above. As a side effect, put_user() will also be affected even though it isn't implicated. Also harden copy_from_user() by redoing the bounds check within the arm_copy_from_user() code, and NULLing the pointer if out of bounds. Acked-by: Mark Rutland Signed-off-by: Russell King

ARM: 8772/1: kprobes: Prohibit kprobes on get_user functions

2018-05-19T10:35:56+00:00

Since do_undefinstr() uses get_user to get the undefined instruction, it can be called before kprobes processes recursive check. This can cause an infinit recursive exception. Prohibit probing on get_user functions. Fixes: 24ba613c9d6c ("ARM kprobes: core code") Signed-off-by: Masami Hiramatsu Cc: stable@vger.kernel.org Signed-off-by: Russell King

Merge branches 'fixes', 'misc', 'sa1111' and 'sa1100-for-next' into for-next

2018-01-21T15:38:10+00:00