Hello, since upstream pm-utils officially stopped development and closes new bug reports, I'm hoping to get this bug fixed here. If the user sets vm/dirty_bytes or vm/dirty_background_bytes, then the kernel interface will reset the values shown in *_ratio to 0 and use the values in *_bytes instead. When pm-utils goes to battery mode, power.d/laptop-mode will write 60 and 40 into *_ratio. This resets the values in *_bytes to 0 instead. When pm-utils returns to AC mode, it writes the previously recorded values from *_ratio back. Now both *_ratio and *_bytes are 0. The kernel interprets this as literal 0, and write performance absolutely TANKS down to sub-megabyte/s. Summary: pm-utils sets vm/dirty_ratio to 60, then back to its original value. If the original value was 0 (from having set dirty_bytes), the system's write performance is now severely degraded. Thanks!
also see https://lwn.net/Articles/572911/ to understand dirty_ values
Maybe migrating too use debian patches would be easier
(In reply to Pacho Ramos from comment #2) > Maybe migrating too use debian patches would be easier Interesting. But the linked issue is not quite the same bug. From what I can tell, it's about pm-utils setting sub-optimal (default) values, thereby overriding user wishes. My bug report is about pm-utils actually trying to restore values it thinks were the user-defined ones, but which aren't, thereby breaking write performance. Still, it's related. I wonder, tho, if the power.d/laptop-mode script is even sensible anymore on modern systems with SSDs, or if the easiest solution would be to just disable it by default.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=f146bb7e14ac3405429772d4adfef555f7810c20 commit f146bb7e14ac3405429772d4adfef555f7810c20 Author: Jakov Smolic <jakov.smolic@sartura.hr> AuthorDate: 2021-02-21 10:56:16 +0000 Commit: David Seifert <soap@gentoo.org> CommitDate: 2021-02-21 10:56:16 +0000 sys-power/pm-utils: Remove last-rited pkg Closes: https://bugs.gentoo.org/659616 Closes: https://bugs.gentoo.org/703026 Closes: https://bugs.gentoo.org/666380 Closes: https://bugs.gentoo.org/579912 Closes: https://bugs.gentoo.org/549848 Closes: https://bugs.gentoo.org/533168 Closes: https://bugs.gentoo.org/489650 Closes: https://bugs.gentoo.org/443530 Closes: https://bugs.gentoo.org/442052 Closes: https://bugs.gentoo.org/437930 Closes: https://bugs.gentoo.org/426952 Closes: https://bugs.gentoo.org/426510 Closes: https://bugs.gentoo.org/403523 Closes: https://bugs.gentoo.org/401257 Closes: https://bugs.gentoo.org/362387 Closes: https://bugs.gentoo.org/257156 Signed-off-by: Jakov Smolic <jakov.smolic@sartura.hr> Signed-off-by: David Seifert <soap@gentoo.org> profiles/arch/amd64/package.use.mask | 4 - profiles/arch/base/package.use.mask | 4 - profiles/arch/hppa/use.mask | 4 - profiles/arch/x86/package.use.mask | 4 - sys-power/pm-utils/Manifest | 1 - .../pm-utils/files/1.4.1-bluetooth-sync.patch | 43 ---------- .../pm-utils/files/1.4.1-disable-sata-alpm.patch | 24 ------ sys-power/pm-utils/files/1.4.1-echo-n.patch | 51 ------------ sys-power/pm-utils/files/1.4.1-fix-alpm-typo.patch | 29 ------- .../1.4.1-fix-intel-audio-powersave-hook.patch | 40 --------- .../pm-utils/files/1.4.1-ignore-led-failure.patch | 14 ---- .../files/1.4.1-inhibit-on-right-status.patch | 20 ----- .../pm-utils/files/1.4.1-logging-append.patch | 17 ---- .../pm-utils/files/1.4.1-run-hook-logging.patch | 17 ---- .../pm-utils/files/1.4.1-suspend-hybrid.patch | 36 --------- .../files/1.4.1-uswsusp-hibernate-mode.patch | 19 ----- .../files/1.4.1-xfs_buffer_arguments.patch | 31 ------- sys-power/pm-utils/files/pm-utils.logrotate | 8 -- sys-power/pm-utils/files/power.d/pci_devices | 61 -------------- sys-power/pm-utils/files/power.d/usb_bluetooth | 45 ----------- sys-power/pm-utils/files/sleep.d/50unload_alx | 18 ----- sys-power/pm-utils/metadata.xml | 10 --- sys-power/pm-utils/pm-utils-1.4.1-r7.ebuild | 94 ---------------------- 23 files changed, 594 deletions(-)