Re: [libvirt] PATCH: User mode linux driver

On Wed, Nov 19, 2008 at 02:49:01PM +0100, Daniel Veillard wrote: > On Wed, Nov 19, 2008 at 01:12:55PM +0000, Daniel P. Berrange wrote: > > This patch is an update of > Hummm, that change to console code parsing seems rather unrelated, > isn't it ? It is needed actually - the old logic was subtley flawed. The original logic was saying if no serial devices were present, and a console was specified then make that console look like a serial device. This is compatability for old mistake where we treated Xen HVM serial console as a paravirt <console/>. This fixup logic was accidentailly being applied to all types of domain, when it should only have been applied to HVM guests. This causes problem with User Mode Linux, where you explicitly have 0 or more paravirt consoles, and 0 or more serial ports. So my change here, just restricts the fixup to hvm only.

> > + for (i = 0 ; i < 16 ; i++) { > > + char *ret; > > + if (i == 0 && vm->def->console) > > + ret = umlBuildCommandLineChr(conn, vm->def->console, "con"); > > + else > > + if (asprintf(&ret, "con%d=none", i) < 0) > > + goto no_memory; > > + ADD_ARG(ret); > > + } > > + > > + for (i = 0 ; i < 16 ; i++) { > > + virDomainChrDefPtr chr = NULL; > > + char *ret; > > + for (j = 0 ; j < vm->def->nserials ; j++) > > + if (vm->def->serials[j]->dstPort == i) > > + chr = vm->def->serials[j]; > > + if (chr) > > + ret = umlBuildCommandLineChr(conn, chr, "ssl"); > > + else > > + if (asprintf(&ret, "ssl%d=none", i) < 0) > > + goto no_memory; > > + ADD_ARG(ret); > > + } > > I'm a bit puzzled by 16 is that an internal UML limit ? Yes, thats a hardcoded user mode linux limit.