PERFORCE change 99406 for review
Divacky Roman
xdivac02 at stud.fit.vutbr.cz
Tue Jun 20 09:18:16 UTC 2006
On Mon, Jun 19, 2006 at 10:02:25AM -0400, John Baldwin wrote:
> On Saturday 17 June 2006 04:57, Roman Divacky wrote:
> > http://perforce.freebsd.org/chv.cgi?CH=99406
> >
> > Change 99406 by rdivacky at rdivacky_witten on 2006/06/17 08:56:59
> >
> > Change the emulpath sysctl to be per-prison.
> >
> > ==== //depot/projects/soc2006/rdivacky_linuxolator/compat/linux/linux_mib.c#3 (text+ko) ====
> >
> > @@ -90,7 +90,7 @@
> > static int
> > linux_sysctl_linux_emul_path(SYSCTL_HANDLER_ARGS)
> > {
> > - char emul_path[LINUX_NAME_MAX];
> > + static char emul_path[LINUX_NAME_MAX];
> > int error;
> > Elf32_Brandinfo **brandinfo;
> >
> > @@ -99,10 +99,39 @@
> > if (error || req->newptr == NULL)
> > return (error);
> > error = linux_set_emul_path(req->td, emul_path);
> > - for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
> > - ++brandinfo)
> > - if (elf32_insert_brand_entry(*brandinfo) < 0)
> > - error = EINVAL;
> > + {
> > + static Elf32_Brandinfo linux_brand = {
> > + ELFOSABI_LINUX,
> > + EM_386,
> > + "Linux",
> > + emul_path,
> > + "/lib/ld-linux.so.1",
> > + &elf_linux_sysvec,
> > + NULL,
> > + BI_CAN_EXEC_DYN,
> > + };
> > + static Elf32_Brandinfo linux_glibc2brand = {
> > + ELFOSABI_LINUX,
> > + EM_386,
> > + "Linux",
> > + emul_path,
> > + "/lib/ld-linux.so.2",
> > + &elf_linux_sysvec,
> > + NULL,
> > + BI_CAN_EXEC_DYN,
> > + };
> > +
> > + Elf32_Brandinfo *local_linux_brandlist[] = {
> > + &linux_brand,
> > + &linux_glibc2brand,
> > + NULL
> > + };
> > +
> > + for (brandinfo = &local_linux_brandlist[0]; *brandinfo != NULL;
> > + ++brandinfo)
> > + if (elf32_insert_brand_entry(*brandinfo) < 0)
> > + error = EINVAL;
> > + }
> > return (error);
> > }
>
> I don't think this is going to work because the brand lists aren't
> per-prison, they are one shared list. That is, once you create a
> jail and try to change the path, the insert's will add new entries
> to the end of the list, but an earlier entry for another jail
> might handle the binary instead. Also, elf32_insert_brand_entry()
hm... so there's no (easy) way to do it per-prison, right?
> saves the pointer you pass to it, and your pointers are for variables
> that are on the stack which will become garbage when this function
> returns.
the variable is declared static so it should work
More information about the p4-projects
mailing list