Re: [RFC v3 08/19] arch: um: add shim to trap to allow installing a fault catcher for tests

From: Brendan Higgins
Date: Mon Dec 03 2018 - 19:45:12 EST


On Mon, Dec 3, 2018 at 3:46 PM Luis Chamberlain <mcgrof@xxxxxxxxxx> wrote:
>
> On Mon, Dec 03, 2018 at 03:34:57PM -0800, Brendan Higgins wrote:
> > On Thu, Nov 29, 2018 at 7:34 PM Luis Chamberlain <mcgrof@xxxxxxxxxx> wrote:
> > >
> > > On Wed, Nov 28, 2018 at 11:36:25AM -0800, Brendan Higgins wrote:
> > > > diff --git a/arch/um/kernel/trap.c b/arch/um/kernel/trap.c
> > > > index cced829460427..bf90e678b3d71 100644
> > > > --- a/arch/um/kernel/trap.c
> > > > +++ b/arch/um/kernel/trap.c
> > > > @@ -201,6 +201,12 @@ void segv_handler(int sig, struct siginfo *unused_si, struct uml_pt_regs *regs)
> > > > segv(*fi, UPT_IP(regs), UPT_IS_USER(regs), regs);
> > > > }
> > > >
> > > > +static void segv_run_catcher(jmp_buf *catcher, void *fault_addr)
> > > > +{
> > > > + current->thread.fault_addr = fault_addr;
> > > > + UML_LONGJMP(catcher, 1);
> > > > +}
> > > > +
> > > > /*
> > > > * We give a *copy* of the faultinfo in the regs to segv.
> > > > * This must be done, since nesting SEGVs could overwrite
> > > > @@ -219,7 +225,10 @@ unsigned long segv(struct faultinfo fi, unsigned long ip, int is_user,
> > > > if (!is_user && regs)
> > > > current->thread.segv_regs = container_of(regs, struct pt_regs, regs);
> > > >
> > > > - if (!is_user && (address >= start_vm) && (address < end_vm)) {
> > > > + catcher = current->thread.fault_catcher;
> > >
> > > This and..
> > >
> > > > + if (catcher && current->thread.is_running_test)
> > > > + segv_run_catcher(catcher, (void *) address);
> > > > + else if (!is_user && (address >= start_vm) && (address < end_vm)) {
> > > > flush_tlb_kernel_vm();
> > > > goto out;
> > > > }
> > >
> > > *not this*
> >
> > I don't understand. Are you saying the previous block of code is good
> > and this one is bad?
>
> No, I was saying that the above block of code is a functional change,
> but I was also pointing out other areas which were not and could be
> folded into a separate atomic patch where no functionality changes.
>
> > > > @@ -246,12 +255,10 @@ unsigned long segv(struct faultinfo fi, unsigned long ip, int is_user,
> > > > address = 0;
> > > > }
> > > >
> > > > - catcher = current->thread.fault_catcher;
> > > > if (!err)
> > > > goto out;
> > > > else if (catcher != NULL) {
> > > > - current->thread.fault_addr = (void *) address;
> > > > - UML_LONGJMP(catcher, 1);
> > > > + segv_run_catcher(catcher, (void *) address);
> > > > }
> > > > else if (current->thread.fault_addr != NULL)
> > > > panic("fault_addr set but no fault catcher");
> > >
> > > But with this seems one atomic change which should be submitted
> > > separately, its just a helper. Think it would make the actual
> > > change needed easier to review, ie, your needed changes would
> > > be smaller and clearer for what you need.
> >
> > Are you suggesting that I pull out the bits needed to implement abort
> > in the next patch and squash it into this one?
>
> No, I'm suggesting you can probably split this patch in 2, one which
> wraps things with no functional changes, and another which adds your
> changes.
>

That makes sense.

Thanks for the clarification!