[Bug 2171415] Re: [Regression] Intel H97 SATA I/O errors and DMAR faults with kernel 7.0, avoided by libata.force=max_sec=2560

I found a similar bug here: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2166838 They mention a test kernel 7.0.0-42 which is available from this PPA: https://launchpad.net/~canonical-kernel-team/+archive/ubuntu/proposed2 Can you please test this kernel and see if this fixes the issue? -- You received this bug notification because you are subscribed to linux in Ubuntu. Matching subscriptions: Bgg, Bmail, Nb https://bugs.launchpad.net/bugs/2171415 Title: [Regression] Intel H97 SATA I/O errors and DMAR faults with kernel 7.0, avoided by libata.force=max_sec=2560 Status in linux package in Ubuntu: New Status in linux source package in Resolute: New Bug description: SUMMARY Ubuntu 26.04 (kernel 7.0.0-30-generic) causes reproducible SATA I/O errors and data corruption on Gigabyte GA-H97-D3H motherboards. The problem disappears when limiting the maximum I/O request size using libata.force=max_sec=2560. HARDWARE Motherboard: Gigabyte GA-H97-D3H, BIOS F5 SATA controller: Intel H97 AHCI (8086:8c82), driver ahci SSD: Crucial MX100 256 GB (also reproduced with Samsung 870 EVO) Filesystem: ext4 on LVM on LUKS REPRODUCTION Boot Ubuntu 26.04 with kernel 7.0.0-30-generic (default parameters). On a disposable ext4 filesystem mounted at /target, run: fio --name=installer-like --filename=/target/var/tmp/fio-test.bin --rw=write --bs=4M --ioengine=sync --direct=0 --size=16G --fsync=16 --end_fsync=1 --group_reporting A/B TEST RESULTS Without workaround: max_sectors_kb = 4096 FAIL after 128 MiB (5.7 seconds) fio reports EIO during fsync. Kernel reports Intel DMAR/IOMMU faults for PCI device 00:1f.2, followed by SATA ICRC/handshake errors, failed WRITE FPDMA QUEUED commands and ext4 I/O errors. SMART UDMA_CRC_Error_Count increased from 37 to 44. With libata.force=max_sec=2560: max_sectors_kb = 1280 PASS: 16 GiB written in 84.7 seconds, no I/O errors. Both tests used the same kernel, hardware, filesystem and fio command. ADDITIONAL OBSERVATIONS - Ubuntu 26.04 installation repeatedly fails on this hardware without the workaround, but completes successfully with it. - Reproduced on three identical Gigabyte motherboards, using different SSDs, SATA cables and power supplies. - Ubuntu 24.04 with kernel 6.8 installs successfully. - Large direct-I/O fio and sequential dd tests did not reproduce the error. - The failing kernel log shows 4 MiB ATA DMA writes and DMAR faults immediately preceding SATA errors. The results suggest a regression involving large I/O requests and possibly Intel IOMMU/DMA mapping. Full dmesg, SMART, lspci and fio logs for both test configurations are attached. To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2171415/+subscriptions