yogev_ezra said: "
problem starts when kernel is already booting."... "it looks like his IDE controller has some bug..."
"He already provided a full Everest report:" download/file.php?id=5965
Here is further he
arsay that the ALi M5229 IDE controller (which acts as member of the ALi M1531/M1533 chipset in)
danielsoys
TI/Acer Extensa 355 notebook, causes problems
: https://forums.gentoo.org/viewtopic-t-3 ... t-dma.html
The ALi M5229 IDE controller which is part of the ALi 1533 and 1543 chipsets is extremely buggy.
Only the M5229 version used in the 1543C (C stands for 'corrected') chipset is safe to use with
UltraDMA transfers, as well as certain revisions of the 1543 chipset. However ALi never made available
a list of which M5229 revisions are safe. From my experience, the revisions 0xC2, 0xC3 and 0xC4 are safe
to use with UDMA. The 0x20 revision is not safe, and expect data corruption and data loss if you try {using} it
{with UDMA}. That's why all Linux kernels after 2.4.17 (if I remember correctly) disable UDMA on the buggy revisions....
(sometimes higher {good} revisions don't get correctly detected, so the driver falls back to assuming revision 0x20).
But, again, this is extremely unsafe... My advice is...
stick to {set to use} Multiword DMA modes on those {good} controllers (the relevant hdparm parameter is -X mdma2,
although the kernel libata should enable that all by itself). It's better to have your data accessed in a slower fashion
than to risk losing your data completely.
(So if I understand above last sentences, his advice in
the general case is to always set an M5229 IDE Controller to
Multiword Mode 2, so it cannot use UDMA modes - but we do not have to take such advice for Linux (and Windows)
which should recognize the good and the bad Revisions.
)
About
danielsoys TI/Acer Extensa 355 notebook;
The Everest.log does not clearly display to me that Win95 was using MultiWord Mode 2 (ATA-2),
as I would have thought that
a DMA assignment should be visible for the channel the Hard Drive is on.
Incidently, the IBM DDLA-21620 Hard Drive that is being used can have up to PIO 4 or MultiWord Mode 2 speed
http://www.hgst.com/tech/techlib.nsf/te ... dla_ps.pdf
http://en.wikipedia.org/wiki/Parallel_A ... A_revision
- so is KolibriOS usable with MultiWord Mode 2? and if not could it be set so e.g. PIO 4 is used instead?