Forums before death by AOL, social media and spammers... "We can't have nice things"
|    linux.debian.kernel    |    Debian kernel discussions    |    2,884 messages    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
|    Message 1,613 of 2,884    |
|    Salvatore Bonaccorso to Filippo Giunchedi    |
|    Bug#1121006: linux: reported optimal_io_    |
|    19 Nov 25 18:10:01    |
      XPost: linux.debian.bugs.dist       From: carnil@debian.org              Control: tags -1 + moreinfo              Hi Filippo,              On Wed, Nov 19, 2025 at 01:46:21PM +0000, Filippo Giunchedi wrote:       > Source: linux       > Version: 6.12.57-1       > Severity: important       >       > Dear Maintainer,       > At Wikimedia Foundation we are running Trixie debian-installer on Dell r450       > hardware with an mpt3sas (HBA355i with id 1000:00e6) controller and SSD       > attached. While debian-installer finished successfully, grub was then unable       to       > boot the installed system.       >       > Partman is instructed to assemble a raid10 over four devices with LVM on top.       > Upon inspection the LVM PV is created with ~4GB metadata area which tricks       grub       > into allocating the same amount of memory during LVM detection. While       > grub-install taking ~4GB of RAM "works" during installation, albeit       > grub-install being quite slow, it obviously fails when booting.       >       > I tracked down the problem to md0 reporting optimal_io_size of ~4GB, and LVM       > defaults to align metadata with said size, resulting in abnormally large       > PV metadata area.       >       > The large md0 optimal_io_size seems to come from component devices reporting       > 16MB optimal_io_size as shown below.       >       > This host was working fine with Bookworm, which makes me think something has       > changed in mpt3sas.       >       > My understanding is that the controller queries devices via block limits VPD       > page for these values, and I'm attaching the output below. The original task       > which spawned this work is https://phabricator.wikimedia.org/T407586       >       > I'm happy to conduct further testing for bug fixes and/or investigation.       >       > best,       > Filippo       >       > ====       >       > # uname -a       > Linux cloudcontrol2010-dev 6.12.57+deb13-amd64 #1 SMP PREEMPT_DYNAMIC Debian       6.12.57-1 (2025-11-05) x86_64 GNU/Linux       >       > # lsblk -t       > NAME ALIGNMENT MIN-IO OPT-IO PHY-SEC LOG-SEC ROTA SCHED        RQ-SIZE RA WSAME       > sda 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > |-sda1 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > |-sda2 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > `-sda3 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > `-md0 0 524288 4293918720 4096 512         4192256 0B       > |-vg0-swap 0 524288 4293918720 4096 512         4192256 0B       > |-vg0-root 0 524288 4293918720 4096 512         4192256 0B       > `-vg0-srv 0 524288 4293918720 4096 512         4192256 0B       > sdb 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > |-sdb1 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > |-sdb2 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > `-sdb3 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > `-md0 0 524288 4293918720 4096 512         4192256 0B       > |-vg0-swap 0 524288 4293918720 4096 512         4192256 0B       > |-vg0-root 0 524288 4293918720 4096 512         4192256 0B       > `-vg0-srv 0 524288 4293918720 4096 512         4192256 0B       > sdc 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > |-sdc1 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > |-sdc2 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > `-sdc3 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > `-md0 0 524288 4293918720 4096 512         4192256 0B       > |-vg0-swap 0 524288 4293918720 4096 512         4192256 0B       > |-vg0-root 0 524288 4293918720 4096 512         4192256 0B       > `-vg0-srv 0 524288 4293918720 4096 512         4192256 0B       > sdd 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > |-sdd1 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > |-sdd2 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > `-sdd3 0 4096 16773120 4096 512 0       mq-deadline 256 32760 0B       > `-md0 0 524288 4293918720 4096 512         4192256 0B       > |-vg0-swap 0 524288 4293918720 4096 512         4192256 0B       > |-vg0-root 0 524288 4293918720 4096 512         4192256 0B       > `-vg0-srv 0 524288 4293918720 4096 512         4192256 0B       >       >       > # pvck --dump headers /dev/md0       > label_header at 512       > label_header.id LABELONE       > label_header.sector 1       > label_header.crc 0xbdf3a961       > label_header.offset 32       > label_header.type LVM2 001       > pv_header at 544       > pv_header.pv_uuid KDkSuWsrIico15Y0PenxiLzT8Ad2dGLa       > pv_header.device_size 1919546294272       > pv_header.disk_locn[0] at 584 # location of data area       > pv_header.disk_locn[0].offset 4293918720       > pv_header.disk_locn[0].size 0       > pv_header.disk_locn[1] at 600 # location list end       > pv_header.disk_locn[1].offset 0       > pv_header.disk_locn[1].size 0       > pv_header.disk_locn[2] at 616 # location of metadata area       > pv_header.disk_locn[2].offset 4096       > pv_header.disk_locn[2].size 4293914624       > pv_header.disk_locn[3] at 632 # location list end       > pv_header.disk_locn[3].offset 0       > pv_header.disk_locn[3].size 0       > pv_header_extension at 648       > pv_header_extension.version 2       > pv_header_extension.flags 1              [continued in next message]              --- SoupGate-Win32 v1.05        * Origin: you cannot sedate... all the things you hate (1:229/2)    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
(c) 1994, bbs@darkrealms.ca