| Re: [hatari-devel] IDE LBA-48 not working |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/hatari-devel Archives
]
Hi Uwe,
On 12.2.2024 22.12, Uwe Seimet wrote:
Has anybody successfully tested LBA-48? As far as I can tell it does not
work.
Thomas ported Qemu IDE hard disk code in (early) 2009 to Hatari. After
that, I think most IDE testing (and hard disk image testing in general)
probably has been done by you...
I created a drive image of about 250 GB and used it for an emulated
SCSI drive. With the SCSI drive there were no issues. Then I used the same
drive image for IDE and noticed that some of the data written by the SCSI
emulation were not present.
I investigated further and found that this appears to be related to the
LBA-48 mode. The last sector where Hatari returned the same data for (i.e.
the same for IDE and for SCSI) was sector 268435455, which is 2^28 - 1. This
is the last sector where LBA-28 still works. For sector 268435456 Hatari IDE
returns data, at least not those this sector actually contains. Hatari does
not even log that the sector was accessed. When writing to this sector it is
similar: There is no error, nothing is logged, but the data are not written.
Maybe it's something that was not fully supported by 2009 Qemu code
ported to Hatari?
Hatari code has e.g. couple of FIXMEs about "HOB" bit handling, which
may be related to LBA-48 mode handling(?).
Looking at Qemu commit history, that was fixed in Qemu only in 2020:
https://github.com/qemu/qemu/commit/be8c9423dec7bd0a0af7f57ecbbcb2718db72555
And there's a fix to LBA-28 / LBA-48 detection in 2021:
https://github.com/qemu/qemu/commit/46e018e9b741731842b93ce23a86fad60445969b
Thomas?
- Eero