[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: New compiler tools release: cris-dist-1.56
Yes, the bug is serious but requires rather unusual circumstances:
1. The mul instruction has to be just before a fetch of a new
cacheline i.e. in the end of a cacheline or in a delay slot
2. The source operand has to be 0xb0XXX258-0xb0XXX25f (where XXX is
an arbitrary number).
3. It only affects kernel mode
So yes, this bug may potentially have occured in the kernel and would
also disappear mysteriously when the kernel is rebuilt with other
Note that the bug does not affect ETRAX 100 (because the multiplication
was added in LX).
Someone else have to answer the rest of your questions.
From: firstname.lastname@example.org">mailto:email@example.com] On
Behalf Of Ronny L Nilsson
Sent: Monday, March 01, 2004 10:58 PM
To: Hans-Peter Nilsson; dev-etrax
Subject: Re: New compiler tools release: cris-dist-1.56
> This release contains a critical workaround for a hardware bug.
> Without the workaround there may be intermittent crashes and
> other failures. From the NEWS file in the cris-dist-1.56.tar.gz
> file (or /usr/doc/cris-dist-1.56/NEWS for the RPM package):
> - There's a workaround in GCC for a hardware bug related to
> multiplication instructions, item 5 in the errata list at>
This look to me like a rather serious bug. Funny it haven't been
reported earlier. Howcome? Is it "rare enough" to live with for
devboard_lx-R2_1_0 systems or should those be updated too (even though
there is no later official than R2.1.0...)?
The above errata seem to have been recently updated, why hasn't there
been any announcement on this list or at the news column at
http://developer.axis.com? (Or have I missed it?)
Last, but not least. The updated errata seem strange to me. The
previous revision v1.3 had as item no. 5 a description of a SDRAM bug.
Where is that now? Can't find it at all in new rev. Also. new and old
(CVS) revision number seem to equals (1.3).