Re: [PATCH v2 2.6.38-rc8-tip 3/20] 3: uprobes: Breakground pagereplacement.

[Srikar Dronamraju]

* Steven Rostedt <rostedt@xxxxxxxxxxx> [2011-03-14 11:38:57]:

> On Mon, 2011-03-14 at 19:04 +0530, Srikar Dronamraju wrote:
> > +/*
> > + * Called with tsk->mm->mmap_sem held (either for read or write and
> > + * with a reference to tsk->mm.
> > + */
> > +static int write_opcode(struct task_struct *tsk, struct uprobe * uprobe,
> > +                       unsigned long vaddr, uprobe_opcode_t opcode)
> > +{
> > +       struct page *old_page, *new_page;
> > +       void *vaddr_old, *vaddr_new;
> > +       struct vm_area_struct *vma;
> > +       spinlock_t *ptl;
> > +       pte_t *orig_pte;
> > +       unsigned long addr;
> > +       int ret = -EINVAL;
> > +
> > +       /* Read the page with vaddr into memory */
> > +       ret = get_user_pages(tsk, tsk->mm, vaddr, 1, 1, 1, &old_page, &vma);
> > +       if (ret <= 0)
> > +               return -EINVAL;
> > +       ret = -EINVAL;
> > +
> > +       /*
> > +        * check if the page we are interested is read-only mapped
> > +        * Since we are interested in text pages, Our pages of interest
> > +        * should be mapped read-only.
> > +        */
> > +       if ((vma->vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)) ==
> > +                                               (VM_READ|VM_EXEC))
> > +               goto put_out;
> > + 
> 
> I'm confused by the above comment and code. You state we are only
> interested text pages mapped read-only, but then if the page is mapped
> read/exec we exit out? It is fine if it is anything but READ/EXEC.

You are right, it should have been
	if ((vma->vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)) !=
					(VM_READ|VM_EXEC))
		goto put_out;


Your comment applied for read_opcode function too.
Will correct in the next version of the patchset.

However in the next patch, where we replace the above with
valid_vma and that does the right thing.

> 
> I'm also curious to why we can't modify text code that is also mapped as
> read/write.
> 

If text code is mapped read/write then on memory pressure the page gets
written to the disk. Hence breakpoints inserted may end up being in the
disk copy modifying the actual copy.

-- 
Thanks and Regards
Srikar
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/