View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0012310 | Rocky-Linux-10 | kernel | public | 2026-03-23 14:53 | 2026-03-23 14:53 |
| Reporter | Leo Bergolth | Assigned To | |||
| Priority | high | Severity | block | Reproducibility | always |
| Status | new | Resolution | open | ||
| Product Version | 10.1 | ||||
| Summary | 0012310: kernels > 6.12.0-68.el10: data loss when using md raid1 with with writemostly flag | ||||
| Description | The current redhat kernels 6.12.0-124.40.1.el10_1 on EL10 and 5.14.0-611.36.1.el9_7 on EL9 contain a regression that leads to data-loss if a raid1 is configured with one disk flagged as writemostly. The regression was introduced by this linux kernel commit: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=8a0adf3d778c4a0893c6d34a9e1b0082a6f1c495 It was backported to EL10 and EL9: EL10: * Mon Mar 24 2025 Julio Faracco <jfaracco@redhat.com> [6.12.0-68.el10] - md/raid1,raid10: don't ignore IO flags (Nigel Croxon) [RHEL-83951] EL9: * Thu Mar 27 2025 Augusto Caringi <acaringi@redhat.com> [5.14.0-576.el9] - md/raid1,raid10: don't ignore IO flags (Nigel Croxon) [RHEL-83988] The mainline kernel already contains a fix: https://github.com/torvalds/linux/commit/743bf030947169c413a711f60cebe73f837e649f However this fix didn't make it into redhat kernels yet. | ||||
| Steps To Reproduce | Originally it happened on a two disk raid1 with one member flagged as writemostly. However I could reproduce it reliably by only connecting the readmostly marked member in degraded mode: - connect the disk - write some data to it (writes will succeed but no data is actually written) - drop caches - re-read data and verify that nothing has been written | ||||
| Tags | No tags attached. | ||||
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2026-03-23 14:53 | Leo Bergolth | New Issue |
