With the built in IDE controller in a Rev. A G3 and a more modern hard disk (like my Maxtor 20gig 2MB buffer dma66 diamondmax) I get 4 lost interrupts a day relating to hdc.
This controller can't do U/DMA 66. All it can do is MDMA2 which shouldn't trigger lost interrupts. Maybe we have something wrong in the DBDMA code ?
According to a note from Andre Hedrick, under some circumstances, on large hard disks (20/30 Gb and more), some timeouts may happen magically due to the time needed for the drive to flush a cache a seek to a block far away from it's current head position. The common IDE code is beeing reworked to handle such timeouts more gracefully, I will adapt pmac-ide when that's done.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Additional note: That was also reported to me with the U/DMA controller of the G4s. The timeouts appear only occasionally but cause DMA to be disabled.
There may still be the possibility of a bug in our DBDMA implementation but I really don't think so.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
On my machine (a beige G3/266 rev 2) at least, this bug is still active as of kernel 2.4.1-pre7_0b on 9-Feb-2001.
Does this still happen ?
This controller can't do U/DMA 66. All it can do is MDMA2 which shouldn't trigger lost interrupts. Maybe we have something wrong in the DBDMA code ?
According to a note from Andre Hedrick, under some circumstances, on large hard disks (20/30 Gb and more), some timeouts may happen magically due to the time needed for the drive to flush a cache a seek to a block far away from it's current head position. The common IDE code is beeing reworked to handle such timeouts more gracefully, I will adapt pmac-ide when that's done.
Additional note: That was also reported to me with the U/DMA controller of the G4s. The timeouts appear only occasionally but cause DMA to be disabled.
There may still be the possibility of a bug in our DBDMA implementation but I really don't think so.
Can you give me more details ? (frequency of the problem, dmesg log when that happens, etc...). Thanks.
Ben.