riscv: Increase mmap_rnd_bits_max on Sv48/57
Subsystems (4)
cover riscv: Increase mmap_rnd_bits_max on Sv48/57
Hi all,
We noticed that 64-bit RISC-V kernels limit mmap_rnd_bits to 24 even if the hardware supports a larger virtual address space size [1]. These two patches allow mmap_rnd_bits_max to be changed during init, and bumps up the maximum randomness if we end up setting up 4/5-level paging at boot.
Sami
[1] https://github.com/google/android-riscv64/issues/1
Sami Tolvanen (2):
mm: Change mmap_rnd_bits_max to __ro_after_init
riscv: mm: Update mmap_rnd_bits_max
arch/riscv/mm/init.c | 6 ++++++ include/linux/mm.h | 2 +- mm/mmap.c | 2 +- 3 files changed, 8 insertions(+), 2 deletions(-)
base-commit: 6465e260f48790807eef06b583b38ca9789b6072
Signature · 1 line
2.42.0.582.g8ccd20d70d-goog
PATCH 1/2 mm: Change mmap_rnd_bits_max to __ro_after_init
Allow mmap_rnd_bits_max to be updated on architectures that determine virtual address space size at runtime instead of relying on Kconfig options by changing it from const to __ro_after_init.
--- include/linux/mm.h | 2 +- mm/mmap.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
include/linux/mm.h +1 -1
@@ -86,7 +86,7 @@ extern int sysctl_legacy_va_layout;#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS extern const int mmap_rnd_bits_min; -extern const int mmap_rnd_bits_max; +extern int mmap_rnd_bits_max__ro_after_init ;extern int mmap_rnd_bits__read_mostly ;#endif #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
mm/mmap.c +1 -1
@@ -64,7 +64,7 @@#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS const int mmap_rnd_bits_min =CONFIG_ARCH_MMAP_RND_BITS_MIN ; -const int mmap_rnd_bits_max =CONFIG_ARCH_MMAP_RND_BITS_MAX ; +int mmap_rnd_bits_max__ro_after_init =CONFIG_ARCH_MMAP_RND_BITS_MAX ;int mmap_rnd_bits__read_mostly =CONFIG_ARCH_MMAP_RND_BITS ;#endif #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
Signature · 1 line
2.42.0.582.g8ccd20d70d-goog
Re: [PATCH 1/2] mm: Change mmap_rnd_bits_max to __ro_after_init
Quoted 5 lines · Sami Tolvanen
Allow
mmap_rnd_bits_maxto be updated on architectures that determine virtual address space size at runtime instead of relying onKconfigoptions by changing it from const to__ro_after_init.Signed-off-by Sami Tolvanen <[email protected]>
Just the kind of thing __ro_after_init was made for. :)
Signature · 1 line
Kees Cook
PATCH 2/2 riscv: mm: Update mmap_rnd_bits_max
In reply to Sami TolvanenARCH_MMAP_RND_BITS_MAX is based on Sv39, which leaves a few potential bits of mmap randomness on the table if we end up enabling 4/5-level paging. Update mmap_rnd_bits_max to take the final address space size into account. This increases mmap_rnd_bits_max from 24 to 33 with Sv48/57.
--- arch/riscv/mm/init.c | 6 ++++++ 1 file changed, 6 insertions(+)
arch/riscv/mm/init.c +6 -0
@@ -762,6 +762,11 @@ static int __init print_no5lvl(char *p) }early_param ("no5lvl" , print_no5lvl); +static void __init set_mmap_rnd_bits_max (void ) +{ + mmap_rnd_bits_max =MMAP_VA_BITS -PAGE_SHIFT -3 ; +} +/* * There is a simple way to determine if 4-level is supported by the * underlying hardware: establish 1:1 mapping in 4-level page table mode
@@ -1071,6 +1076,7 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa)#if defined(CONFIG_64BIT ) && !defined(CONFIG_XIP_KERNEL )set_satp_mode (dtb_pa); +set_mmap_rnd_bits_max ();#endif /*
Signature · 1 line
2.42.0.582.g8ccd20d70d-goog
Re: [PATCH 2/2] riscv: mm: Update mmap_rnd_bits_max
Quoted 7 lines · Sami Tolvanen
ARCH_MMAP_RND_BITS_MAXis based on Sv39, which leaves a few potential bits of mmap randomness on the table if we end up enabling 4/5-level paging. Updatemmap_rnd_bits_maxto take the final address space size into account. This increasesmmap_rnd_bits_maxfrom 24 to 33 with Sv48/57.Signed-off-by Sami Tolvanen <[email protected]>
I like this. Is RISCV the only arch where the paging level can be chosen at boot time?
Signature · 1 line
Kees Cook
Re: [PATCH 2/2] riscv: mm: Update mmap_rnd_bits_max
Quoted 8 lines · Kees Cook
Quoted 6 lines · Sami Tolvanen
ARCH_MMAP_RND_BITS_MAXis based on Sv39, which leaves a few potential bits of mmap randomness on the table if we end up enabling 4/5-level paging. Updatemmap_rnd_bits_maxto take the final address space size into account. This increasesmmap_rnd_bits_maxfrom 24 to 33 with Sv48/57.
Quoted 4 lines
Quoted 1 line
Signed-off-by Sami Tolvanen <[email protected]>I like this. Is RISCV the only arch where the paging level can be chosen at boot time?
I haven't seen this elsewhere, but I also haven't looked at all the other architectures that closely. arm64 does something interesting with ARM64_VA_BITS_52, but I think we can still handle that in Kconfig.
Sami
Re: [PATCH 2/2] riscv: mm: Update mmap_rnd_bits_max
Quoted 11 lines · Sami Tolvanen
Quoted 10 lines · Kees Cook
Quoted 7 lines · Sami Tolvanen
ARCH_MMAP_RND_BITS_MAXis based on Sv39, which leaves a few potential bits of mmap randomness on the table if we end up enabling 4/5-level paging. Updatemmap_rnd_bits_maxto take the final address space size into account. This increasesmmap_rnd_bits_maxfrom 24 to 33 with Sv48/57.Signed-off-by Sami Tolvanen <[email protected]>
Quoted 7 lines
Quoted 2 lines
I like this. Is RISCV the only arch where the paging level can be chosen at boot time?
I haven't seen this elsewhere, but I also haven't looked at all the other architectures that closely. arm64 does something interesting with
ARM64_VA_BITS_52, but I think we can still handle that inKconfig.
AFAIU, x86-64 can do this also:
no4lvl [RISCV] Disable 4-level and 5-level paging modes. Forces
kernel to use 3-level paging instead.
no5lvl [X86-64,RISCV] Disable 5-level paging mode. Forces
kernel to use 4-level paging instead.
Re: [PATCH 2/2] riscv: mm: Update mmap_rnd_bits_max
Quoted 20 lines · Conor Dooley
Quoted 11 lines · Sami Tolvanen
Quoted 10 lines · Kees Cook
Quoted 7 lines · Sami Tolvanen
ARCH_MMAP_RND_BITS_MAXis based on Sv39, which leaves a few potential bits of mmap randomness on the table if we end up enabling 4/5-level paging. Updatemmap_rnd_bits_maxto take the final address space size into account. This increasesmmap_rnd_bits_maxfrom 24 to 33 with Sv48/57.Signed-off-by Sami Tolvanen <[email protected]>Quoted 7 lines
Quoted 2 lines
I like this. Is RISCV the only arch where the paging level can be chosen at boot time?
I haven't seen this elsewhere, but I also haven't looked at all the other architectures that closely. arm64 does something interesting with
ARM64_VA_BITS_52, but I think we can still handle that inKconfig.
Quoted 7 lines
AFAIU, x86-64 can do this also:
no4lvl [RISCV] Disable 4-level and 5-level paging modes. Forces
kernel to use 3-level paging instead.no5lvl [X86-64,RISCV] Disable 5-level paging mode. Forces
kernel to use 4-level paging instead.
Ah-ha! Okay, well, then let's track this idea:
https://github.com/KSPP/linux/issues/346
Signature · 1 line
Kees Cook
Re: [PATCH 2/2] riscv: mm: Update mmap_rnd_bits_max
Quoted 24 lines · Kees Cook
Quoted 22 lines · Conor Dooley
Quoted 11 lines · Sami Tolvanen
Quoted 10 lines · Kees Cook
Quoted 7 lines · Sami Tolvanen
ARCH_MMAP_RND_BITS_MAXis based on Sv39, which leaves a few potential bits of mmap randomness on the table if we end up enabling 4/5-level paging. Updatemmap_rnd_bits_maxto take the final address space size into account. This increasesmmap_rnd_bits_maxfrom 24 to 33 with Sv48/57.Signed-off-by Sami Tolvanen <[email protected]>Quoted 7 lines
Quoted 2 lines
I like this. Is RISCV the only arch where the paging level can be chosen at boot time?
I haven't seen this elsewhere, but I also haven't looked at all the other architectures that closely. arm64 does something interesting with
ARM64_VA_BITS_52, but I think we can still handle that inKconfig.AFAIU, x86-64 can do this also:
Quoted 8 lines
Quoted 5 lines
no4lvl [RISCV] Disable 4-level and 5-level paging modes. Forces
kernel to use 3-level paging instead.no5lvl [X86-64,RISCV] Disable 5-level paging mode. Forces
kernel to use 4-level paging instead.Ah-ha! Okay, well, then let's track this idea:
https://github.com/KSPP/linux/issues/346
(Replying here for visibility, tell me if you want to move this discussion to github)
AIUI, x86 cannot do this for compat reasons. Even if you enable LA57, mmap only gives you < 48-bit addresses, for compatibility with things like JITs, etc that stash information in the upper 16 bits. You need to pass a > 48-bit mmap hint to get 57-bit addresses.
I imagine riscv does not have this issue yet, due to little accumulated cruft, but it may be wise to check against popular JITters for these problems on riscv code.
Signature · 1 line
Pedro
Alexandre Ghiti
reviewerRe: [PATCH 2/2] riscv: mm: Update mmap_rnd_bits_max
Quoted 30 lines · Pedro Falcato
Quoted 19 lines · Kees Cook
Quoted 18 lines · Conor Dooley
Quoted 8 lines · Sami Tolvanen
Quoted 7 lines · Kees Cook
Quoted 6 lines · Sami Tolvanen
ARCH_MMAP_RND_BITS_MAXis based on Sv39, which leaves a few potential bits of mmap randomness on the table if we end up enabling 4/5-level paging. Updatemmap_rnd_bits_maxto take the final address space size into account. This increasesmmap_rnd_bits_maxfrom 24 to 33 with Sv48/57.Quoted 7 lines
Quoted 3 lines
Quoted 1 line
Signed-off-by Sami Tolvanen <[email protected]>I like this. Is RISCV the only arch where the paging level can be chosen at boot time?
I haven't seen this elsewhere, but I also haven't looked at all the other architectures that closely. arm64 does something interesting with
ARM64_VA_BITS_52, but I think we can still handle that inKconfig.AFAIU, x86-64 can do this also:
Quoted 7 lines
Quoted 5 lines
no4lvl [RISCV] Disable 4-level and 5-level paging modes. Forces
kernel to use 3-level paging instead.no5lvl [X86-64,RISCV] Disable 5-level paging mode. Forces
kernel to use 4-level paging instead.Ah-ha! Okay, well, then let's track this idea:
https://github.com/KSPP/linux/issues/346(Replying here for visibility, tell me if you want to move this discussion to github)
Quoted 9 lines
AIUI, x86 cannot do this for compat reasons. Even if you enable LA57, mmap only gives you < 48-bit addresses, for compatibility with things like JITs, etc that stash information in the upper 16 bits. You need to pass a > 48-bit mmap hint to get 57-bit addresses.
I imagine riscv does not have this issue yet, due to little accumulated cruft, but it may be wise to check against popular JITters for these problems on riscv code.
We already encountered those issues and the same solution was recently merged (restrict to sv48 unless otherwise specified): https://lore.kernel.org/all/[email protected]/
Re: [PATCH 2/2] riscv: mm: Update mmap_rnd_bits_max
Quoted 37 lines · Alexandre Ghiti
Quoted 35 lines · Pedro Falcato
Quoted 19 lines · Kees Cook
Quoted 18 lines · Conor Dooley
Quoted 8 lines · Sami Tolvanen
Quoted 7 lines · Kees Cook
Quoted 6 lines · Sami Tolvanen
ARCH_MMAP_RND_BITS_MAXis based on Sv39, which leaves a few potential bits of mmap randomness on the table if we end up enabling 4/5-level paging. Updatemmap_rnd_bits_maxto take the final address space size into account. This increasesmmap_rnd_bits_maxfrom 24 to 33 with Sv48/57.Quoted 7 lines
Quoted 3 lines
Quoted 1 line
Signed-off-by Sami Tolvanen <[email protected]>I like this. Is RISCV the only arch where the paging level can be chosen at boot time?
I haven't seen this elsewhere, but I also haven't looked at all the other architectures that closely. arm64 does something interesting with
ARM64_VA_BITS_52, but I think we can still handle that inKconfig.AFAIU, x86-64 can do this also:
Quoted 7 lines
Quoted 5 lines
no4lvl [RISCV] Disable 4-level and 5-level paging modes. Forces
kernel to use 3-level paging instead.no5lvl [X86-64,RISCV] Disable 5-level paging mode. Forces
kernel to use 4-level paging instead.Ah-ha! Okay, well, then let's track this idea:
https://github.com/KSPP/linux/issues/346(Replying here for visibility, tell me if you want to move this discussion to github)
AIUI, x86 cannot do this for compat reasons. Even if you enable LA57, mmap only gives you < 48-bit addresses, for compatibility with things like JITs, etc that stash information in the upper 16 bits. You need to pass a > 48-bit mmap hint to get 57-bit addresses.
Quoted 8 lines
Quoted 4 lines
I imagine riscv does not have this issue yet, due to little accumulated cruft, but it may be wise to check against popular JITters for these problems on riscv code.
We already encountered those issues and the same solution was recently merged (restrict to sv48 unless otherwise specified):
https://lore.kernel.org/all/[email protected]/
We recently ran into this issue when bringing up Android as well because qemu defaults to Sv57 and some userspace bits weren't happy with >48-bit mmap addresses.
Note that this patch uses MMAP_VA_BITS, which is 48 for both Sv48 and Sv57, which is why mmap_rnd_bits_max will be 33 even with Sv57.
Sami
Quoted 7 lines · On Fri, 29 Sep 2023 14:11:56 PDT (-0700), [email protected] wrote:
Hi all,
We noticed that 64-bit RISC-V kernels limit
mmap_rnd_bitsto 24 even if the hardware supports a larger virtual address space size [1]. These two patches allowmmap_rnd_bits_maxto be changed during init, and bumps up the maximum randomness if we end up setting up 4/5-level paging at boot.
Sorry for missing this, I'm just poking through old stuff in patchwork. As far as I can tell this is still relevant, the discussions are just on the mmap() bits (but we'd already screwed that one up and have since fixed it).
So
in case someone else wants to take it, but I'm OK taking that MM patch with Kees' review.
Quoted 17 lines
Sami
[1] https://github.com/google/android-riscv64/issues/1
Sami Tolvanen (2):
mm: Changemmap_rnd_bits_maxto__ro_after_init
riscv: mm: Updatemmap_rnd_bits_maxarch/riscv/mm/init.c | 6 ++++++ include/linux/mm.h | 2 +- mm/mmap.c | 2 +- 3 files changed, 8 insertions(+), 2 deletions(-)base-commit:
6465e260f48790807eef06b583b38ca9789b6072
Quoted 13 lines · Palmer Dabbelt
Quoted 7 lines · On Fri, 29 Sep 2023 14:11:56 PDT (-0700), [email protected] wrote:
Hi all,
We noticed that 64-bit RISC-V kernels limit
mmap_rnd_bitsto 24 even if the hardware supports a larger virtual address space size [1]. These two patches allowmmap_rnd_bits_maxto be changed during init, and bumps up the maximum randomness if we end up setting up 4/5-level paging at boot.Sorry for missing this, I'm just poking through old stuff in patchwork. As far as I can tell this is still relevant, the discussions are just on the
mmap()bits (but we'd already screwed that one up and have since fixed it).
Quoted 7 lines
So
Reviewed-by Palmer Dabbelt <[email protected]>Acked-by Palmer Dabbelt <[email protected]>in case someone else wants to take it, but I'm OK taking that MM patch with Kees' review.
Yes, thanks! Please do. I already +1ed it:
https://lore.kernel.org/all/202309291454.436E19663@keescook
-Kees
Signature · 1 line
Kees Cook
Hi Palmer,
Quoted 15 lines · Palmer Dabbelt
Quoted 7 lines · On Fri, 29 Sep 2023 14:11:56 PDT (-0700), [email protected] wrote:
Hi all,
We noticed that 64-bit RISC-V kernels limit
mmap_rnd_bitsto 24 even if the hardware supports a larger virtual address space size [1]. These two patches allowmmap_rnd_bits_maxto be changed during init, and bumps up the maximum randomness if we end up setting up 4/5-level paging at boot.Sorry for missing this, I'm just poking through old stuff in patchwork. As far as I can tell this is still relevant, the discussions are just on the
mmap()bits (but we'd already screwed that one up and have since fixed it).
Quoted 7 lines
So
Reviewed-by Palmer Dabbelt <[email protected]>Acked-by Palmer Dabbelt <[email protected]>in case someone else wants to take it, but I'm OK taking that MM patch with Kees' review.
Is this still on your radar for v6.8 or would you prefer me to resend the patches?
Sami
This series was applied to riscv/linux.git (for-next) by Palmer Dabbelt <[email protected]>:
Hello:
This series was applied to riscv/linux.git (for-next) by Palmer Dabbelt <[email protected]>:
Quoted 9 lines · On Fri, 29 Sep 2023 21:11:56 +0000 you wrote:
Hi all,
We noticed that 64-bit RISC-V kernels limit
mmap_rnd_bitsto 24 even if the hardware supports a larger virtual address space size [1]. These two patches allowmmap_rnd_bits_maxto be changed during init, and bumps up the maximum randomness if we end up setting up 4/5-level paging at boot.[...]
Here is the summary with links:
- [1/2] mm: Change mmap_rnd_bits_max to __ro_after_init
https://git.kernel.org/riscv/c/71a5849aedaa - [2/2] riscv: mm: Update mmap_rnd_bits_max https://git.kernel.org/riscv/c/7df1ff5a5cd6
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html