| Message ID | 20221109045112.187069-1-bgray@linux.ibm.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | powerpc/code-patching: Use temporary mm for Radix MMU | expand |
On Wed, 9 Nov 2022 15:51:03 +1100, Benjamin Gray wrote: > This is a revision of Chris and Jordan's series to introduce a per-cpu temporary > mm to be used for patching with strict rwx on radix mmus. > > v10: * Don't set poking_init_enabled on boot CPU init failure > * Remove extern from functions in files touched by this series > * Changed book3s/32/tlbflush.h stub body from warning to build bug > * Remove empty hash__ set of functions. Also removed the ones that are > just warnings (would have preferred to make them build bugs, but they > blocked the build even though it's all inline inside a radix_enabled() > if-block). > - It's not possible to use the radix specific necessary TLB flush > in code patching, as code-patching.c is compiled even when radix__* > functions aren't visible. #ifdefs would be required. > The cxl usage required radix__* be visible anyway, so there is no > compile issue. > * Mention cache benefits of struct of patching context variables > * Don't open-code mm init and teardown. It seems tlb_gather_mmu does > the necessary steps to detect page table pages to free, so it should > not leak. get_locked_pte() is used over __get_locked_pte() because I > don't know what makes the double-underscore function dangerous. > > [...] Patches 1-2 and 4-9 applied to powerpc/next. [1/9] powerpc: Allow clearing and restoring registers independent of saved breakpoint state https://git.kernel.org/powerpc/c/3671f4ebe3eb12e7222e4d7b0f94e85cfe34253a [2/9] powerpc/code-patching: Use WARN_ON and fix check in poking_init https://git.kernel.org/powerpc/c/071c95c1acbd96e76bab8b25b5cad0d71a011f37 [4/9] powerpc/mm: Remove empty hash__ functions https://git.kernel.org/powerpc/c/baf1ed24b27db475b38f534953885d0425e2232d [5/9] cxl: Use radix__flush_all_mm instead of generic flush_all_mm https://git.kernel.org/powerpc/c/0f0a0a6091e678b1a75078ecd6b02176f3228dbb [6/9] powerpc/mm: Remove flush_all_mm, local_flush_all_mm https://git.kernel.org/powerpc/c/d34471c9bd5d47ab148dd68817631a4238f755c4 [7/9] powerpc/tlb: Add local flush for page given mm_struct and psize https://git.kernel.org/powerpc/c/274d842fa1efd9449e62222c8896e0be11621f1f [8/9] powerpc/code-patching: Use temporary mm for Radix MMU https://git.kernel.org/powerpc/c/c28c15b6d28a776538482101522cbcd9f906b15c [9/9] powerpc/code-patching: Consolidate and cache per-cpu patching context https://git.kernel.org/powerpc/c/2f228ee1ade5d8d1f26cf94863a36c5693023c58 cheers
This is a revision of Chris and Jordan's series to introduce a per-cpu temporary mm to be used for patching with strict rwx on radix mmus. v10: * Don't set poking_init_enabled on boot CPU init failure * Remove extern from functions in files touched by this series * Changed book3s/32/tlbflush.h stub body from warning to build bug * Remove empty hash__ set of functions. Also removed the ones that are just warnings (would have preferred to make them build bugs, but they blocked the build even though it's all inline inside a radix_enabled() if-block). - It's not possible to use the radix specific necessary TLB flush in code patching, as code-patching.c is compiled even when radix__* functions aren't visible. #ifdefs would be required. The cxl usage required radix__* be visible anyway, so there is no compile issue. * Mention cache benefits of struct of patching context variables * Don't open-code mm init and teardown. It seems tlb_gather_mmu does the necessary steps to detect page table pages to free, so it should not leak. get_locked_pte() is used over __get_locked_pte() because I don't know what makes the double-underscore function dangerous. Previous versions: v9: https://lore.kernel.org/all/20221025044409.448755-1-bgray@linux.ibm.com/ v8: https://lore.kernel.org/all/20221021052238.580986-1-bgray@linux.ibm.com/ v7: https://lore.kernel.org/all/20211110003717.1150965-1-jniethe5@gmail.com/ v6: https://lore.kernel.org/all/20210911022904.30962-1-cmr@bluescreens.de/ v5: https://lore.kernel.org/all/20210713053113.4632-1-cmr@linux.ibm.com/ v4: https://lore.kernel.org/all/20210429072057.8870-1-cmr@bluescreens.de/ v3: https://lore.kernel.org/all/20200827052659.24922-1-cmr@codefail.de/ v2: https://lore.kernel.org/all/20200709040316.12789-1-cmr@informatik.wtf/ v1: https://lore.kernel.org/all/20200603051912.23296-1-cmr@informatik.wtf/ RFC: https://lore.kernel.org/all/20200323045205.20314-1-cmr@informatik.wtf/ x86: https://lore.kernel.org/kernel-hardening/20190426232303.28381-1-nadav.amit@gmail.com/ Benjamin Gray (9): powerpc: Allow clearing and restoring registers independent of saved breakpoint state powerpc/code-patching: Use WARN_ON and fix check in poking_init powerpc/mm: Remove extern from function prototypes powerpc/mm: Remove empty hash__ functions cxl: Use radix__flush_all_mm instead of generic flush_all_mm powerpc/mm: Remove flush_all_mm, local_flush_all_mm powerpc/tlb: Add local flush for page given mm_struct and psize powerpc/code-patching: Use temporary mm for Radix MMU powerpc/code-patching: Consolidate and cache per-cpu patching context arch/powerpc/include/asm/book3s/32/tlbflush.h | 9 + .../include/asm/book3s/64/tlbflush-hash.h | 69 +----- arch/powerpc/include/asm/book3s/64/tlbflush.h | 41 +--- arch/powerpc/include/asm/debug.h | 2 + arch/powerpc/include/asm/mmu_context.h | 6 +- arch/powerpc/include/asm/nohash/tlbflush.h | 27 ++- arch/powerpc/kernel/process.c | 38 +++- arch/powerpc/lib/code-patching.c | 207 ++++++++++++++++-- arch/powerpc/mm/nohash/tlb.c | 8 + 9 files changed, 287 insertions(+), 120 deletions(-) base-commit: 247f34f7b80357943234f93f247a1ae6b6c3a740 -- 2.38.1