Hi Ben, I was following this discussion as I was affected from the very same issue as well. Axel Beckert was helping me, analyzing this issue further (thank you, Axel).
I installed the test kernel you mentioned in your last email and it worked for me. In my case I encountered issues with an Areca controller (Areca Technology Corp. ARC-188x series PCIe 2.0/3.0 to SAS/SATA 6/12Gb RAID Controller (rev 02)) that was not recognized anymore. I only saw the following errors in dmesg: [Thu Nov 28 00:33:11 2024] Areca RAID Controller0: Model ARC-1883, F/W V1.70 2024-08-05 [Thu Nov 28 00:33:11 2024] arcmsr0: dma_alloc_coherent got error [Thu Nov 28 00:33:11 2024] arcmsr0: arcmsr_alloc_ccb_pool got error With your test kernel the correct output is now: [ 4.599360] Areca RAID Controller0: Model ARC-1883, F/W V1.70 2024-08-05 [ 4.607260] scsi host0: Areca SAS/SATA RAID Controller (RAID6 capable) arcmsr version v1.50.00.05-20210429 [ 4.608990] arcmsr0: msi-x enabled [ 4.635797] arcmsr0: cdb_phyaddr_hi32=0x11 [ 4.991584] scsi 0:0:0:0: Direct-Access Areca ARC-1883-VOL#000 R001 PQ: 0 ANSI: 5 [ 4.994960] scsi 0:0:16:0: Processor Areca RAID controller R001 PQ: 0 ANSI: 0 Thank you very much for your work. I appreciate it a lot! Best wishes, David On Wednesday, 19 February 2025 at 21:18:41 +01:00, Ben Hutchings <b...@decadent.org.uk> wrote: > Hi Nicolas, > > Sorrry for the delay in handling this bug. > > We think this regression is due to the upstream change: > > commit 9f40ec84a7976d95c34e7cc070939deb103652b0 > Author: Juergen Gross <<jgr...@suse.com>> > Date: Fri Sep 13 12:05:02 2024 +0200 > > xen/swiotlb: add alignment check for dma buffers > > which was backported to the 6.1-stable series. > > To confirm this, please can you test: > > 1. Is this fixed in the latest released kernel for bookworm (version > 6.1.128-1)? If it is, you can stop here. > > 2. Is this fixed in version 6.1.128-1a~test from > <<https://people.debian.org/~benh/packages/linux-xen-regression/>>? > (This reverts the above change.) > > 3. Is this fixed in version 6.1.128-1a~test2, from the same place? > (This adds an upstream fix on top of the above change.) > > Note that the test packages above are not signed for use with Secure > Boot. > > Ben. > > -- > Ben Hutchings > Klipstein's 4th Law of Prototyping and Production: > A fail-safe circuit will destroy others. >
