Messages in this thread | | | Date | Thu, 24 Nov 2011 19:55:07 +0530 | From | Srikar Dronamraju <> | Subject | Re: [PATCH v7 3.2-rc2 4/30] uprobes: Define hooks for mmap/munmap. |
| |
* Peter Zijlstra <peterz@infradead.org> [2011-11-24 15:13:37]:
> On Thu, 2011-11-24 at 19:17 +0530, Srikar Dronamraju wrote: > > * Peter Zijlstra <peterz@infradead.org> [2011-11-23 19:10:12]: > > > > > On Fri, 2011-11-18 at 16:37 +0530, Srikar Dronamraju wrote: > > > > + ret = install_breakpoint(vma->vm_mm, uprobe); > > > > + if (ret == -EEXIST) { > > > > + atomic_inc(&vma->vm_mm->mm_uprobes_count); > > > > + ret = 0; > > > > + } > > > > > > Aren't you double counting that probe position here? The one that raced > > > you to inserting it will also have incremented that counter, no? > > > > > > > No we arent. > > Because register_uprobe can never race with mmap_uprobe and register > > before mmap_uprobe registers .(Once we start mmap_region, > > register_uprobe waits for the read_lock of mmap_sem.) > > Still doesn't make any sense. Since you don't increment on success, one > has to assume install_breakpoint() will cause an increment. Therefore, > when we encounter -EEXIST we'll already have accounted for this > mm,inode,offset combination. >
In the success case, install_breakpoint itself does the increment. We cant allow install_breakpoint to increment in EEXIST case always because doing that in register_uprobe context would increment which is wrong.
> But I'll have another look at it, maybe I'm missing something > obvious :-) > > > And we badly need this for mmap_uprobe case. Because when we do mremap, > > or vma_adjust(), we do a munmap_uprobe() followed by mmap_uprobe() which > > would have decremented the count but not removed it. So when we do a > > mmap_uprobe, we need to increment the count. > > Well I see why the count needs to be correct, that's not the issue.
Okay ..
-- Thanks and Regards Srikar
| |