[PATCH] xc_ptrace.c fix

[PATCH] xc_ptrace.c fix

[Kip Macy]

Could you please apply the following trivial patch (yes I know, gmail
screws up whitespace). It fixes a crash in the gdbserver that can be
hit when debugging linux.


--- a/tools/libxc/xc_ptrace.c   Fri Jul  1 21:25:45 2005
+++ b/tools/libxc/xc_ptrace.c   Sun Jul  3 01:05:21 2005
@@ -200,7 +200,7 @@
                                             pde_phys[cpu] >>
PAGE_SHIFT)) == NULL)
            goto error_out;
     }
-    if ((page = pde_virt[cpu][vtopti(va)]) == 0) /* logical address */
+    if ((pde_virt[cpu] == NULL) || (page = pde_virt[cpu][vtopti(va)])
== 0) /* logical address */
        goto error_out;
     if (ctxt[cpu].flags & VGCF_VMX_GUEST)
         page = page_array[page >> PAGE_SHIFT] << PAGE_SHIFT;

Re: [PATCH] xc_ptrace.c fix

[Keir Fraser]

On 3 Jul 2005, at 02:07, Kip Macy wrote:

> Could you please apply the following trivial patch (yes I know, gmail
> screws up whitespace). It fixes a crash in the gdbserver that can be
> hit when debugging linux.

But you set pde_virt immediately above the test that you are patching, 
if pde_phys[cpu] != pde.
And if mmap'ing pde_virt fails, you bail with an error. So how can you 
get pde_virt[cpu] == NULL at that point -- aren;t you just hiding the 
real bug?

  -- Keir