Control: notfound -1 linux/5.10.70-1 On Monday, 30 May 2022 22:54:07 CEST Diederik de Haas wrote: > Control: found -1 linux/5.10.84-1 > Control: retitle -1 SATA hotplug on AMD x570 MB broken since 5.10.0-10 > > On Tue, 17 May 2022 14:25:29 +0100 Bob <s...@homeurl.co.uk> wrote: > > On my AMD x570 chipset based system the kernel linux-image-5.10.0-10-amd64 > > broke SATA hotplug, if I revert to linux-image-5.10.0-9-amd64 hotplug > > works,
linux-image-5.10.0-9 corresponds to Debian version 5.10.70-1; so update metadata accordingly The best way to figure out which commit broke SATA hotplug, is doing a `git bisect` operations. Normally that means building a 5.10.70 (upstream) kernel and verifying it does work with that and then a 5.10.84 (upstream) kernel and verifying it breaks with that version and then starting the `git bisect` operation with that. Verifying with those 2 upstream kernel versions (still) sounds like the best start. But you can (hopefully) speed that up by making some educated guesses. For that I've filtered the commits in `drivers/ata` between those version and looked up the git tags between them: ~/dev/kernel.org/linux$ git log --oneline v5.10.70..v5.10.84 -- drivers/ata/ v5.10.84 16c242b09106 sata_fsl: fix warning in remove_proc_entry when rmmod sata_fsl 4a46b2f5dce0 sata_fsl: fix UAF in sata_fsl_port_stop when rmmod sata_fsl 9a32d3c08d8e ata: ahci: Add Green Sardine vendor ID as board_ahci_mobile v5.10.83 v5.10.80 701cf28e019d libata: fix checking of DMA state 39264eaa6d44 libata: fix read log timeout value v5.10.79 v5.10.77 5aa5bab57957 ata: sata_mv: Fix the error handling of mv_chip_id() v5.10.76 v5.10.75 7eef482db728 ata: ahci_platform: fix null-ptr-deref in ahci_platform_enable_regulators() 116932c0e45e pata_legacy: fix a couple uninitialized variable bugs v5.10.74 v5.10.72 387aecdab7fa libata: Add ATA_HORKAGE_NO_NCQ_ON_ATI for Samsung 860 and 870 SSD. v5.10.71 My recommendation would be to try an upstream kernel v5.10.76 and if that works try the v5.10.80 version. If .76 works and .80 fails then you'd have a narrower set to `git bisect` and should be able to find the offending commit sooner. You could also try 5.10.79 as an extra 'educated guess'. Do you know how to build an upstream kernel and how to do `git bisect`?
signature.asc
Description: This is a digitally signed message part.
