Masked user access avoids the address/size verification by access_ok().
Allthough its main purpose is to skip the speculation in the
verification of user address and size hence avoid the need of spec
mitigation, it also has the advantage of reducing the amount of
instructions required so it even benefits to platforms that don't
need speculation mitigation, especially when the size of the copy is
not know at build time.
So implement masked user access on powerpc. The only requirement is
to have memory gap that faults between the top user space and the
real start of kernel area.
On 64 bits platforms the address space is divided that way:
0xffffffffffffffff +------------------+
| |
| kernel space |
| |
0xc000000000000000 +------------------+ <== PAGE_OFFSET
|//////////////////|
|//////////////////|
0x8000000000000000 |//////////////////|
|//////////////////|
|//////////////////|
0x0010000000000000 +------------------+ <== TASK_SIZE_MAX
| |
| user space |
| |
0x0000000000000000 +------------------+
Kernel is always above 0x8000000000000000 and user always
below, with a gap in-between. It leads to a 3 instructions sequence:
150: 7c 69 fe 76 sradi r9,r3,63
154: 79 29 00 40 clrldi r9,r9,1
158: 7c 63 48 78 andc r3,r3,r9
This sequence leaves r3 unmodified when it is below 0x8000000000000000
and clamps it to 0x8000000000000000 if it is above.
On 32 bits it is more tricky. In theory user space can go up to
0xbfffffff while kernel will usually start at 0xc0000000. So a gap
needs to be added in-between. Allthough in theory a single 4k page
would suffice, it is easier and more efficient to enforce a 128k gap
below kernel, as it simplifies the masking.
e500 has the isel instruction which allows selecting one value or
the other without branch and that instruction is not speculative, so
use it. Allthough GCC usually generates code using that instruction,
it is safer to use inline assembly to be sure. The result is:
14: 3d 20 bf fe lis r9,-16386
18: 7c 03 48 40 cmplw r3,r9
1c: 7c 69 18 5e iselgt r3,r9,r3
On other ones, when kernel space is over 0x80000000 and user space
is below, the logic in mask_user_address_simple() leads to a
3 instruction sequence:
64: 7c 69 fe 70 srawi r9,r3,31
68: 55 29 00 7e clrlwi r9,r9,1
6c: 7c 63 48 78 andc r3,r3,r9
This is the default on powerpc 8xx.
When the limit between user space and kernel space is not 0x80000000,
mask_user_address_32() is used and a 6 instructions sequence is
generated:
24: 54 69 7c 7e srwi r9,r3,17
28: 21 29 57 ff subfic r9,r9,22527
2c: 7d 29 fe 70 srawi r9,r9,31
30: 75 2a b0 00 andis. r10,r9,45056
34: 7c 63 48 78 andc r3,r3,r9
38: 7c 63 53 78 or r3,r3,r10
The constraint is that TASK_SIZE be aligned to 128K in order to get
the most optimal number of instructions.
When CONFIG_PPC_BARRIER_NOSPEC is not defined, fallback on the
test-based masking as it is quicker than the 6 instructions sequence
but not quicker than the 3 instructions sequences above.
As an exemple, allthough barrier_nospec() voids on the 8xx, this
change has the following impact on strncpy_from_user(): the length of
the function is reduced from 488 to 340 bytes:
Start of the function with the patch:
00000000 <strncpy_from_user>:
0: 7c ab 2b 79 mr. r11,r5
4: 40 81 01 40 ble 144 <strncpy_from_user+0x144>
8: 7c 89 fe 70 srawi r9,r4,31
c: 55 29 00 7e clrlwi r9,r9,1
10: 7c 84 48 78 andc r4,r4,r9
14: 3d 20 dc 00 lis r9,-9216
18: 7d 3a c3 a6 mtspr 794,r9
1c: 2f 8b 00 03 cmpwi cr7,r11,3
20: 40 9d 00 b4 ble cr7,d4 <strncpy_from_user+0xd4>
...
Start of the function without the patch:
00000000 <strncpy_from_user>:
0: 7c a0 2b 79 mr. r0,r5
4: 40 81 01 10 ble 114 <strncpy_from_user+0x114>
8: 2f 84 00 00 cmpwi cr7,r4,0
c: 41 9c 01 30 blt cr7,13c <strncpy_from_user+0x13c>
10: 3d 20 80 00 lis r9,-32768
14: 7d 24 48 50 subf r9,r4,r9
18: 7f 80 48 40 cmplw cr7,r0,r9
1c: 7c 05 03 78 mr r5,r0
20: 41 9d 01 00 bgt cr7,120 <strncpy_from_user+0x120>
24: 3d 20 80 00 lis r9,-32768
28: 7d 25 48 50 subf r9,r5,r9
2c: 7f 84 48 40 cmplw cr7,r4,r9
30: 38 e0 ff f2 li r7,-14
34: 41 9d 00 e4 bgt cr7,118 <strncpy_from_user+0x118>
38: 94 21 ff e0 stwu r1,-32(r1)
3c: 3d 20 dc 00 lis r9,-9216
40: 7d 3a c3 a6 mtspr 794,r9
44: 2b 85 00 03 cmplwi cr7,r5,3
48: 40 9d 01 6c ble cr7,1b4 <strncpy_from_user+0x1b4>
...
118: 7c e3 3b 78 mr r3,r7
11c: 4e 80 00 20 blr
120: 7d 25 4b 78 mr r5,r9
124: 3d 20 80 00 lis r9,-32768
128: 7d 25 48 50 subf r9,r5,r9
12c: 7f 84 48 40 cmplw cr7,r4,r9
130: 38 e0 ff f2 li r7,-14
134: 41 bd ff e4 bgt cr7,118 <strncpy_from_user+0x118>
138: 4b ff ff 00 b 38 <strncpy_from_user+0x38>
13c: 38 e0 ff f2 li r7,-14
140: 4b ff ff d8 b 118 <strncpy_from_user+0x118>
...
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Madhavan Srinivasan <maddy@linux.ibm.com>
Link: https://patch.msgid.link/8f418183d9125cc0bf23922bc2ef2a1130d8b63a.1766574657.git.chleroy@kernel.org
At the time being, TASK_SIZE can be customized by the user via Kconfig
but it is not possible to check all constraints in Kconfig. Impossible
setups are detected at compile time with BUILD_BUG() but that leads
to build failure when setting crazy values. It is not a problem on its
own because the user will usually either use the default value or set
a well thought value. However build robots generate crazy random
configs that lead to build failures, and build robots see it as a
regression every time a patch adds such a constraint.
So instead of failing the build when the custom TASK_SIZE is too
big, just adjust it to the maximum possible value matching the setup.
Several architectures already calculate TASK_SIZE based on other
parameters and options.
In order to do so, move MODULES_VADDR calculation into task_size_32.h
and ensure that:
- On book3s/32, userspace and module area have their own segments (256M)
- On 8xx, userspace has its own full PGDIR entries (4M)
Then TASK_SIZE is guaranteed to be correct so remove related
BUILD_BUG()s.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Madhavan Srinivasan <maddy@linux.ibm.com>
Link: https://patch.msgid.link/6a2575420770d075cd090b5a316730a2ffafdee4.1766574657.git.chleroy@kernel.org
For book3s/32 it is assumed that TASK_SIZE is a multiple of 256 Mbytes,
but Kconfig allows any value for TASK_SIZE.
In all relevant calculations, align TASK_SIZE to the upper 256 Mbytes
boundary.
Also use ASM_CONST() in the definition of TASK_SIZE to ensure it is
seen as an unsigned constant.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Madhavan Srinivasan <maddy@linux.ibm.com>
Link: https://patch.msgid.link/8928d906079e156c59794c41e826a684eaaaebb4.1766574657.git.chleroy@kernel.org
When activating CONFIG_THREAD_INFO_IN_TASK, linux/sched.h includes
asm/current.h. This generates a circular dependency. To avoid that,
asm/processor.h shall not be included in mmu-hash.h.
In order to do that, this patch moves into a new header called
asm/task_size_64/32.h all the TASK_SIZE related constants, which can
then be included in mmu-hash.h directly.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Split out all the TASK_SIZE constants not just 64-bit ones]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>