Message ID | 20210802215408.804942-1-pasha.tatashin@soleen.com (mailing list archive) |
---|---|
Headers | show |
Series | arm64: MMU enabled kexec relocation | expand |
Hi Pavel, This series is still missing reviews from those who understand kexec better than me. On Mon, Aug 02, 2021 at 05:53:53PM -0400, Pavel Tatashin wrote: > Enable MMU during kexec relocation in order to improve reboot performance. > > If kexec functionality is used for a fast system update, with a minimal > downtime, the relocation of kernel + initramfs takes a significant portion > of reboot. > > The reason for slow relocation is because it is done without MMU, and thus > not benefiting from D-Cache. The performance improvements are indeed significant on some platforms (going from 7s to ~40ms), so I think the merging the series is worth it. Some general questions so I better understand the impact: - Is the kdump path affected in any way? IIUC that doesn't need any relocation but we should also make sure we don't create the additional page table unnecessarily (should keep as much memory intact as possible). Maybe that's already handled. - What happens if trans_pgd_create_copy() fails to allocate memory. Does it fall back to an MMU-off relocation? And I presume this series does not introduce any changes to the kexec tools ABI. Thanks.
On Tue, Aug 24, 2021 at 2:06 PM Catalin Marinas <catalin.marinas@arm.com> wrote: > > Hi Pavel, > > This series is still missing reviews from those who understand kexec > better than me. Hi Catalin, Yes, I am looking for reviewers. > > On Mon, Aug 02, 2021 at 05:53:53PM -0400, Pavel Tatashin wrote: > > Enable MMU during kexec relocation in order to improve reboot performance. > > > > If kexec functionality is used for a fast system update, with a minimal > > downtime, the relocation of kernel + initramfs takes a significant portion > > of reboot. > > > > The reason for slow relocation is because it is done without MMU, and thus > > not benefiting from D-Cache. > > The performance improvements are indeed significant on some platforms > (going from 7s to ~40ms), so I think the merging the series is worth it. > Some general questions so I better understand the impact: > > - Is the kdump path affected in any way? IIUC that doesn't need any > relocation but we should also make sure we don't create the additional > page table unnecessarily (should keep as much memory intact as > possible). Maybe that's already handled. Because kdump does not need relocation, we do not reserve pages for the page table in the kdump reboot case. In fact, with this series, kdump reboot becomes more straightforward as we skip the relocation function entirely, and jump directly into the crash kernel (or purgatory if kexec tools loaded them). > > - What happens if trans_pgd_create_copy() fails to allocate memory. Does > it fall back to an MMU-off relocation? In case we are so low on memory that trans_pgd_create_copy() fails to allocate the linear map that uses the large pages (the size of the page table is tiny) the kexec fails during kexec load time (not during reboot time), as out of memory. The MMU enabled kexec reboot is always on, and we should not have several ways to do kexec reboot as it makes the kexec reboot unpredictable in terms of performance, and also prone to bugs by having a common MMU enabled path and less common path when we are low on memory which is never tested. > > And I presume this series does not introduce any changes to the kexec > tools ABI. Correct. Thanks for taking a look at this series. Pasha
On Mon, Aug 02, 2021 at 05:53:53PM -0400, Pavel Tatashin wrote: > Changelog: > v16: > - Merged with 5.14-rc4 > v15: > - Changed trans_pgd_copy_el2_vectors() to use vector table that > only shared by kexec and hibernate. This way sync does not have > dangling branch that was recently introduced. (Reported by Marc > Zyngier) > - Renamed is_hyp_callable() to is_hyp_nvhe() as requested by Marc > Zyngier > - Clean-ups, comment fixes. > - Sync with upstream 368094df48e680fa51cedb68537408cfa64b788e > v14: > - Fixed a bug in "arm64: hyp-stub: Move elx_sync into the vectors" > that was noticed by Marc Zyngier > - Merged with upstream > v13: > - Fixed a hang on ThunderX2, thank you Pingfan Liu for reporting > the problem. In relocation function we need civac not ivac, we > need to clean data in addition to invalidating it. > Since I was using ThunderX2 machine I also measured the new > performance data on this large ARM64 server. The MMU improves > kexec relocation 190 times on this machine! (see below for > raw data). Saves 7.5s during CentOS kexec reboot. > v12: > - A major change compared to previous version. Instead of using > contiguous VA range a copy of linear map is now used to perform > copying of segments during relocation as it was agreed in the > discussion of version 11 of this project. > - In addition to using linear map, I also took several ideas from > James Morse to better organize the kexec relocation: > 1. skip relocation function entirely if that is not needed > 2. remove the PoC flushing function since it is not needed > anymore with MMU enabled. > v11: > - Fixed missing KEXEC_CORE dependency for trans_pgd.c > - Removed useless "if(rc) return rc" statement (thank you Tyler Hicks) > - Another 12 patches were accepted into maintainer's get. > Re-based patches against: > https://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux.git > Branch: for-next/kexec > v10: > - Addressed a lot of comments form James Morse and from Marc Zyngier > - Added review-by's > - Synchronized with mainline > > v9: - 9 patches from previous series landed in upstream, so now series > is smaller > - Added two patches from James Morse to address idmap issues for machines > with high physical addresses. > - Addressed comments from Selin Dag about compiling issues. He also tested > my series and got similar performance results: ~60 ms instead of ~580 ms > with an initramfs size of ~120MB. > v8: > - Synced with mainline to keep series up-to-date > v7: > -- Addressed comments from James Morse > - arm64: hibernate: pass the allocated pgdp to ttbr0 > Removed "Fixes" tag, and added Added Reviewed-by: James Morse > - arm64: hibernate: check pgd table allocation > Sent out as a standalone patch so it can be sent to stable > Series applies on mainline + this patch > - arm64: hibernate: add trans_pgd public functions > Remove second allocation of tmp_pg_dir in swsusp_arch_resume > Added Reviewed-by: James Morse <james.morse@arm.com> > - arm64: kexec: move relocation function setup and clean up > Fixed typo in commit log > Changed kern_reloc to phys_addr_t types. > Added explanation why kern_reloc is needed. > Split into four patches: > arm64: kexec: make dtb_mem always enabled > arm64: kexec: remove unnecessary debug prints > arm64: kexec: call kexec_image_info only once > arm64: kexec: move relocation function setup > - arm64: kexec: add expandable argument to relocation function > Changed types of new arguments from unsigned long to phys_addr_t. > Changed offset prefix to KEXEC_* > Split into four patches: > arm64: kexec: cpu_soft_restart change argument types > arm64: kexec: arm64_relocate_new_kernel clean-ups > arm64: kexec: arm64_relocate_new_kernel don't use x0 as temp > arm64: kexec: add expandable argument to relocation function > - arm64: kexec: configure trans_pgd page table for kexec > Added invalid entries into EL2 vector table > Removed KEXEC_EL2_VECTOR_TABLE_SIZE and KEXEC_EL2_VECTOR_TABLE_OFFSET > Copy relocation functions and table into separate pages > Changed types in kern_reloc_arg. > Split into three patches: > arm64: kexec: offset for relocation function > arm64: kexec: kexec EL2 vectors > arm64: kexec: configure trans_pgd page table for kexec > - arm64: kexec: enable MMU during kexec relocation > Split into two patches: > arm64: kexec: enable MMU during kexec relocation > arm64: kexec: remove head from relocation argument > v6: > - Sync with mainline tip > - Added Acked's from Dave Young > v5: > - Addressed comments from Matthias Brugger: added review-by's, improved > comments, and made cleanups to swsusp_arch_resume() in addition to > create_safe_exec_page(). > - Synced with mainline tip. > v4: > - Addressed comments from James Morse. > - Split "check pgd table allocation" into two patches, and moved to > the beginning of series for simpler backport of the fixes. > Added "Fixes:" tags to commit logs. > - Changed "arm64, hibernate:" to "arm64: hibernate:" > - Added Reviewed-by's > - Moved "add PUD_SECT_RDONLY" earlier in series to be with other > clean-ups > - Added "Derived from:" to arch/arm64/mm/trans_pgd.c > - Removed "flags" from trans_info > - Changed .trans_alloc_page assumption to return zeroed page. > - Simplify changes to trans_pgd_map_page(), by keeping the old > code. > - Simplify changes to trans_pgd_create_copy, by keeping the old > code. > - Removed: "add trans_pgd_create_empty" > - replace init_mm with NULL, and keep using non "__" version of > populate functions. > v3: > - Split changes to create_safe_exec_page() into several patches for > easier review as request by Mark Rutland. This is why this series > has 3 more patches. > - Renamed trans_table to tans_pgd as agreed with Mark. The header > comment in trans_pgd.c explains that trans stands for > transitional page tables. Meaning they are used in transition > between two kernels. > v2: > - Fixed hibernate bug reported by James Morse > - Addressed comments from James Morse: > * More incremental changes to trans_table > * Removed TRANS_FORCEMAP > * Added kexec reboot data for image with 380M in size. > > Enable MMU during kexec relocation in order to improve reboot performance. > > If kexec functionality is used for a fast system update, with a minimal > downtime, the relocation of kernel + initramfs takes a significant portion > of reboot. > > The reason for slow relocation is because it is done without MMU, and thus > not benefiting from D-Cache. > > Performance data > ---------------- > > Cavium ThunderX2: > Kernel Image size: 38M Iniramfs size: 46M Total relocation size: 84M > MMU-disabled: > relocation 7.489539915s > MMU-enabled: > relocation 0.03946095s > > Relocation performance is improved 190 times. > > Broadcom Stingray: > For this experiment, the size of kernel plus initramfs is small, only 25M. > If initramfs was larger, than the improvements would be greater, as time > spent in relocation is proportional to the size of relocation. > > MMU-disabled:: > kernel shutdown 0.022131328s > relocation 0.440510736s > kernel startup 0.294706768s > > Relocation was taking: 58.2% of reboot time > > MMU-enabled: > kernel shutdown 0.032066576s > relocation 0.022158152s > kernel startup 0.296055880s > > Now: Relocation takes 6.3% of reboot time > > Total reboot is x2.16 times faster. > > With bigger userland (fitImage 380M), the reboot time is improved by 3.57s, > and is reduced from 3.9s down to 0.33s > > Previous approaches and discussions > ----------------------------------- > v15: https://lore.kernel.org/lkml/20210609004419.936873-1-pasha.tatashin@soleen.com > v14: https://lore.kernel.org/lkml/20210527150526.271941-1-pasha.tatashin@soleen.com > v13: https://lore.kernel.org/lkml/20210408040537.2703241-1-pasha.tatashin@soleen.com > v12: https://lore.kernel.org/lkml/20210303002230.1083176-1-pasha.tatashin@soleen.com > v11: https://lore.kernel.org/lkml/20210127172706.617195-1-pasha.tatashin@soleen.com > v10: https://lore.kernel.org/linux-arm-kernel/20210125191923.1060122-1-pasha.tatashin@soleen.com > v9: https://lore.kernel.org/lkml/20200326032420.27220-1-pasha.tatashin@soleen.com > v8: https://lore.kernel.org/lkml/20191204155938.2279686-1-pasha.tatashin@soleen.com > v7: https://lore.kernel.org/lkml/20191016200034.1342308-1-pasha.tatashin@soleen.com > v6: https://lore.kernel.org/lkml/20191004185234.31471-1-pasha.tatashin@soleen.com > v5: https://lore.kernel.org/lkml/20190923203427.294286-1-pasha.tatashin@soleen.com > v4: https://lore.kernel.org/lkml/20190909181221.309510-1-pasha.tatashin@soleen.com > v3: https://lore.kernel.org/lkml/20190821183204.23576-1-pasha.tatashin@soleen.com > v2: https://lore.kernel.org/lkml/20190817024629.26611-1-pasha.tatashin@soleen.com > v1: https://lore.kernel.org/lkml/20190801152439.11363-1-pasha.tatashin@soleen.com > > Pavel Tatashin (15): > arm64: kernel: add helper for booted at EL2 and not VHE > arm64: trans_pgd: hibernate: Add trans_pgd_copy_el2_vectors > arm64: hibernate: abstract ttrb0 setup function > arm64: kexec: flush image and lists during kexec load time > arm64: kexec: skip relocation code for inplace kexec > arm64: kexec: Use dcache ops macros instead of open-coding > arm64: kexec: pass kimage as the only argument to relocation function > arm64: kexec: configure EL2 vectors for kexec > arm64: kexec: relocate in EL1 mode > arm64: kexec: use ld script for relocation function > arm64: kexec: install a copy of the linear-map > arm64: kexec: keep MMU enabled during kexec relocation > arm64: kexec: remove the pre-kexec PoC maintenance > arm64: kexec: remove cpu-reset.h > arm64: trans_pgd: remove trans_pgd_map_page() > > arch/arm64/Kconfig | 2 +- > arch/arm64/include/asm/assembler.h | 49 ++++++-- > arch/arm64/include/asm/kexec.h | 12 ++ > arch/arm64/include/asm/mmu_context.h | 24 ++++ > arch/arm64/include/asm/sections.h | 1 + > arch/arm64/include/asm/trans_pgd.h | 12 +- > arch/arm64/include/asm/virt.h | 7 ++ > arch/arm64/kernel/asm-offsets.c | 11 ++ > arch/arm64/kernel/cpu-reset.S | 7 +- > arch/arm64/kernel/cpu-reset.h | 32 ----- > arch/arm64/kernel/hibernate-asm.S | 72 ----------- > arch/arm64/kernel/hibernate.c | 49 ++------ > arch/arm64/kernel/machine_kexec.c | 177 ++++++++++++++------------- > arch/arm64/kernel/relocate_kernel.S | 70 +++++------ > arch/arm64/kernel/sdei.c | 2 +- > arch/arm64/kernel/vmlinux.lds.S | 19 +++ > arch/arm64/mm/Makefile | 1 + > arch/arm64/mm/trans_pgd-asm.S | 65 ++++++++++ > arch/arm64/mm/trans_pgd.c | 82 ++++--------- > 19 files changed, 356 insertions(+), 338 deletions(-) > delete mode 100644 arch/arm64/kernel/cpu-reset.h > create mode 100644 arch/arm64/mm/trans_pgd-asm.S > > > base-commit: c500bee1c5b2f1d59b1081ac879d73268ab0ff17 > -- Tested-by: Pingfan Liu <piliu@redhat.com>
On Thu, Aug 26, 2021 at 11:03:21AM -0400, Pavel Tatashin wrote: > On Tue, Aug 24, 2021 at 2:06 PM Catalin Marinas <catalin.marinas@arm.com> wrote: > > > Enable MMU during kexec relocation in order to improve reboot performance. > > > > > > If kexec functionality is used for a fast system update, with a minimal > > > downtime, the relocation of kernel + initramfs takes a significant portion > > > of reboot. > > > > > > The reason for slow relocation is because it is done without MMU, and thus > > > not benefiting from D-Cache. > > > > The performance improvements are indeed significant on some platforms > > (going from 7s to ~40ms), so I think the merging the series is worth it. > > Some general questions so I better understand the impact: > > > > - Is the kdump path affected in any way? IIUC that doesn't need any > > relocation but we should also make sure we don't create the additional > > page table unnecessarily (should keep as much memory intact as > > possible). Maybe that's already handled. > > Because kdump does not need relocation, we do not reserve pages for > the page table in the kdump reboot case. In fact, with this series, > kdump reboot becomes more straightforward as we skip the relocation > function entirely, and jump directly into the crash kernel (or > purgatory if kexec tools loaded them). > > > - What happens if trans_pgd_create_copy() fails to allocate memory. Does > > it fall back to an MMU-off relocation? > > In case we are so low on memory that trans_pgd_create_copy() fails to > allocate the linear map that uses the large pages (the size of the > page table is tiny) the kexec fails during kexec load time (not during > reboot time), as out of memory. The MMU enabled kexec reboot is always > on, and we should not have several ways to do kexec reboot as it makes > the kexec reboot unpredictable in terms of performance, and also prone > to bugs by having a common MMU enabled path and less common path when > we are low on memory which is never tested. I think this makes sense, especially since it will fail during the kexec load time rather than reboot. I'm ok in principle with this series but I'd need to convince James Morse to have a another look since he followed it more closely than me. Could you please rebase it against 5.15-rc1? Thanks.
> > In case we are so low on memory that trans_pgd_create_copy() fails to > > allocate the linear map that uses the large pages (the size of the > > page table is tiny) the kexec fails during kexec load time (not during > > reboot time), as out of memory. The MMU enabled kexec reboot is always > > on, and we should not have several ways to do kexec reboot as it makes > > the kexec reboot unpredictable in terms of performance, and also prone > > to bugs by having a common MMU enabled path and less common path when > > we are low on memory which is never tested. > > I think this makes sense, especially since it will fail during the kexec > load time rather than reboot. > > I'm ok in principle with this series but I'd need to convince James > Morse to have a another look since he followed it more closely than me. > Could you please rebase it against 5.15-rc1? Sure, I will send it out soon. Thanks, Pasha