Hello Philip, On Wed, Jan 08, 2025 at 12:52:50PM +0000, Philip Pemberton wrote: > I'm trying to connect an old Iomega Zip 100 ATAPI to a B550-chipset Ryzen > system, to exchange files with an even older system. The Gigabyte B550 AORUS > ELITE AX V2 rev1.3 motherboard doesn't have any PATA ports, so I'm using a > SATA to PATA adapter. > > Sadly it will not work in the B550 system (Kernel 6.8.0-51-generic x86_64, > Linux Mint 21.3 based on Ubuntu 22.04). When I have the Zip drive connected, > I get the following in dmesg and the sd device never appears: > > ata3: SATA link up 1.5 Gbps (SStatus 113 SControl 300) > ata3.00: ATAPI: IOMEGA ZIP 100 ATAPI, 12.A, max PIO3, CDB intr, > DMADIR > ata3.00: applying bridge limits > ata3.00: configured for PIO0 > ata3.00: qc timeout after 5000 msecs (cmd 0xa0) > ata3.00: failed to clear UNIT ATTENTION (err_mask=0x5) > > I've included a more complete dmesg at the bottom of this message. > > I currently have the following in the kernel command line: > libata.atapi_dmadir=1 > libata.force=3.00:atapi_dmadir,dump_id,nodmalog,noncq,pio0 > > I started with only having the DMADIR option as suggested in this old patch > from LKML: > https://lkml.org/lkml/2013/6/18/933 > > With "atapi_dmadir=1" and DMADIR forced, I have the same messages in the > kernel log - except obviously none of the other "horkage" messages or the > ATA IDENTIFY dump, and I think the "xfer mask" starts at a higher speed. > > > The BIOS can see and access the Zip drive fine, as can Windows. > > I've also tried the same setup (SATA bridge) in a Pentium 4 PCI+AGP machine > I had sitting around. Admittedly this isn't much of a test as it was running > a much older and 32bit OS (Knoppix 7.2, kernel 3.9.6) but the sd device > appeared and the drive could be accessed fine. Did you specify anything on the kernel command line when using kernel 3.9.6 ? FWIW, commit e771451c0a83 ("libata: make ata_exec_internal_sg honor DMADIR") was first included in v3.10.1, so even if you did specify it on the kernel command line on kernel 3.9.6, it wouldn't have any effect on internal commands (e.g. IDENTIFY). Kind regards, Niklas
