| Message ID | 20200221160126.GB19330@willie-the-truck (mailing list archive) |
|---|---|
| State | Mainlined |
| Commit | 63f01d852c75366fb4d15ce217d12c48b69a4bcc |
| Headers | show |
| Series | [GIT,PULL] arm64 fixes for -rc3 | expand |
On Fri, Feb 21, 2020 at 8:01 AM Will Deacon <will@kernel.org> wrote: > > That said, this is core code and I know you'd prefer to limit the change > to brk(), so the patch is sitting on top of the branch in case you prefer > not to include it. If you decide to tweak it manually, please can you > update the docs at the same time? I've taken it as-is, since I don't think any of the solutions were wrong. The mremap() case looks odd, with the _old_ address untagged, but the new one not. I see the logic, but because it looks so odd I think it might be worth a comment. Not a big deal, and more of a "if you get around to it later" note. Linus
The pull request you sent on Fri, 21 Feb 2020 16:01:26 +0000:
> git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux.git tags/arm64-fixes
has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/63f01d852c75366fb4d15ce217d12c48b69a4bcc
Thank you!
Hi Linus, Please pull these arm64 fixes for -rc3. It's all straightforward apart from the changes to mmap()/mremap() in relation to their handling of address arguments from userspace with non-zero tag bits in the upper byte. The change to brk() is necessary to fix a nasty user-visible regression in malloc(), but we tightened up mmap() and mremap() at the same time because they also allow the user to create virtual aliases by accident. It's much less likely than brk() to matter in practice, but enforcing the principle of "don't permit the creation of mappings using tagged addresses" leads to a straightforward ABI without having to worry about the "but what if a crazy program did foo?" aspect of things. That said, this is core code and I know you'd prefer to limit the change to brk(), so the patch is sitting on top of the branch in case you prefer not to include it. If you decide to tweak it manually, please can you update the docs at the same time? Cheers, Will --->8 The following changes since commit d91771848f0ae2eec250a9345926a1a3558fa943: arm64: time: Replace <linux/clk-provider.h> by <linux/of_clk.h> (2020-02-12 17:26:38 +0000) are available in the Git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux.git tags/arm64-fixes for you to fetch changes up to dcde237319e626d1ec3c9d8b7613032f0fd4663a: mm: Avoid creating virtual address aliases in brk()/mmap()/mremap() (2020-02-20 10:03:14 +0000) ---------------------------------------------------------------- arm64 fixes for -rc3 - Fix regression in malloc() caused by ignored address tags in brk() - Add missing brackets around argument to untagged_addr() macro - Fix clang build when using binutils assembler - Fix silly typo in virtual memory map documentation ---------------------------------------------------------------- Catalin Marinas (1): mm: Avoid creating virtual address aliases in brk()/mmap()/mremap() Scott Branden (1): docs: arm64: fix trivial spelling enought to enough in memory.rst Vincenzo Frascino (1): arm64: lse: Fix LSE atomics with LLVM Will Deacon (1): arm64: memory: Add missing brackets to untagged_addr() macro Documentation/arm64/memory.rst | 2 +- Documentation/arm64/tagged-address-abi.rst | 11 +++++++++-- arch/arm64/include/asm/lse.h | 2 +- arch/arm64/include/asm/memory.h | 2 +- mm/mmap.c | 4 ---- mm/mremap.c | 1 - 6 files changed, 12 insertions(+), 10 deletions(-)