From: Jim C. <ji...@ma...> - 2005-12-21 07:16:33
|
On Mon, 19 Dec 2005, Blaisorblade wrote: > On Monday 19 December 2005 21:19, Jim Carter wrote: > The problem lies in the #ifdef mapping ntohll to _bswap64 - in 2.6.13 it was > changed in a wrong way (and not by anybody present here - somebody went with > a "nice cleanup" without understanding what was going on). (Groan.) > While the comment refers to a "2.6.13" format and an "other kernels" format, > it's clear that uml_utilities use the latter. > > Also, normally there's no need to use uml_mkcow - passing > ubd0=<cow_path,backing file path> auto setups the COW file if needed (i.e. it > doesn't exist). Unfortunately that was the first thing I tried: the 2.6.13 kernel wrote the same header that uml_mkcow did, then complained that the size was wrong in the header it had just written. It looks like the provided combination of kernel and utilities is not going to be useable; either the kernel has to be repatched or the utilities have to be sabotaged. I'm having a lot of trouble to figure out just which patch SuSE used. That's why I mentioned printing out a patch title when the UML boots, separately for the host and the guest, in case of different versions. James F. Carter Voice 310 825 2897 FAX 310 206 6673 UCLA-Mathnet; 6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA 90095-1555 Email: ji...@ma... http://www.math.ucla.edu/~jimc (q.v. for PGP key) |