| Message ID | 20191126004407.4b72ef7f@suzdal.zaitcev.lan (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | usb: mon: Fix a deadlock in usbmon between mmap and read | expand |
On Tue, 26 Nov 2019, Pete Zaitcev wrote: > The problem arises because our read() function grabs a lock of the > circular buffer, finds something of interest, then invokes copy_to_user() > straight from the buffer, which in turn takes mm->mmap_sem. In the same > time, the callback mon_bin_vma_fault() is invoked under mm->mmap_sem. > It attempts to take the fetch lock and deadlocks. > > This patch does away with protecting of our page list with any > semaphores, and instead relies on the kernel not close the device > while mmap is active in a process. > > In addition, we prohibit re-sizing of a buffer while mmap is active. > This way, when (now unlocked) fault is processed, it works with the > page that is intended to be mapped-in, and not some other random page. > Note that this may have an ABI impact, but hopefully no legitimate > program is this wrong. > > Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> > Reported-by: syzbot+56f9673bb4cdcbeb0e92@syzkaller.appspotmail.com Reviewed-by: Alan Stern <stern@rowland.harvard.edu> Also this should have: Fixes: 46eb14a6e158 ("USB: fix usbmon BUG trigger") CC: <stable@vger.kernel.org> Alan Stern
On Tue, 26 Nov 2019 10:20:14 -0500 (EST) Alan Stern <stern@rowland.harvard.edu> wrote: > > Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> > > Reported-by: syzbot+56f9673bb4cdcbeb0e92@syzkaller.appspotmail.com > > Reviewed-by: Alan Stern <stern@rowland.harvard.edu> Thanks. > Fixes: 46eb14a6e158 ("USB: fix usbmon BUG trigger") Indeed... Either I didn't think that one through, or the copy_to_user used not to take the mmap_sem. > CC: <stable@vger.kernel.org> Do we really need this? The problem was in the code for more than 10 years. It's not like anyone is exploiting systems because of it. If we do need it, I should cc: the submission to the same place too, right? -- Pete
On Tue, Nov 26, 2019 at 10:35:09PM -0600, Pete Zaitcev wrote: > On Tue, 26 Nov 2019 10:20:14 -0500 (EST) > Alan Stern <stern@rowland.harvard.edu> wrote: > > > > Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> > > > Reported-by: syzbot+56f9673bb4cdcbeb0e92@syzkaller.appspotmail.com > > > > Reviewed-by: Alan Stern <stern@rowland.harvard.edu> > > Thanks. > > > Fixes: 46eb14a6e158 ("USB: fix usbmon BUG trigger") > > Indeed... Either I didn't think that one through, or the copy_to_user > used not to take the mmap_sem. > > > CC: <stable@vger.kernel.org> > > Do we really need this? The problem was in the code for more than 10 years. > It's not like anyone is exploiting systems because of it. Well now we all have a simple reproducer for it, so yes, it should be backported. I'm doing that for all of the syzbot stuff. > If we do need it, I should cc: the submission to the same place too, right? Nope, the tag is just fine, that's all that is needed. I'll add the above to the patch when applying it to my trees. thanks, greg k-h
On Tue, 26 Nov 2019, Pete Zaitcev wrote: > On Tue, 26 Nov 2019 10:20:14 -0500 (EST) > Alan Stern <stern@rowland.harvard.edu> wrote: > > > > Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> > > > Reported-by: syzbot+56f9673bb4cdcbeb0e92@syzkaller.appspotmail.com > > > > Reviewed-by: Alan Stern <stern@rowland.harvard.edu> > > Thanks. > > > Fixes: 46eb14a6e158 ("USB: fix usbmon BUG trigger") > > Indeed... Either I didn't think that one through, or the copy_to_user > used not to take the mmap_sem. copy_to_user doesn't, but the fault handler does (the core handler, not the fault routine in mon_bin.c). After all, it doesn't want the memory map to change while a page is being read in to satisfy the fault. Alan Stern
diff --git a/drivers/usb/mon/mon_bin.c b/drivers/usb/mon/mon_bin.c index ac2b4fcc265f..f48a23adbc35 100644 --- a/drivers/usb/mon/mon_bin.c +++ b/drivers/usb/mon/mon_bin.c @@ -1039,12 +1039,18 @@ static long mon_bin_ioctl(struct file *file, unsigned int cmd, unsigned long arg mutex_lock(&rp->fetch_lock); spin_lock_irqsave(&rp->b_lock, flags); - mon_free_buff(rp->b_vec, rp->b_size/CHUNK_SIZE); - kfree(rp->b_vec); - rp->b_vec = vec; - rp->b_size = size; - rp->b_read = rp->b_in = rp->b_out = rp->b_cnt = 0; - rp->cnt_lost = 0; + if (rp->mmap_active) { + mon_free_buff(vec, size/CHUNK_SIZE); + kfree(vec); + ret = -EBUSY; + } else { + mon_free_buff(rp->b_vec, rp->b_size/CHUNK_SIZE); + kfree(rp->b_vec); + rp->b_vec = vec; + rp->b_size = size; + rp->b_read = rp->b_in = rp->b_out = rp->b_cnt = 0; + rp->cnt_lost = 0; + } spin_unlock_irqrestore(&rp->b_lock, flags); mutex_unlock(&rp->fetch_lock); } @@ -1216,13 +1222,21 @@ mon_bin_poll(struct file *file, struct poll_table_struct *wait) static void mon_bin_vma_open(struct vm_area_struct *vma) { struct mon_reader_bin *rp = vma->vm_private_data; + unsigned long flags; + + spin_lock_irqsave(&rp->b_lock, flags); rp->mmap_active++; + spin_unlock_irqrestore(&rp->b_lock, flags); } static void mon_bin_vma_close(struct vm_area_struct *vma) { + unsigned long flags; + struct mon_reader_bin *rp = vma->vm_private_data; + spin_lock_irqsave(&rp->b_lock, flags); rp->mmap_active--; + spin_unlock_irqrestore(&rp->b_lock, flags); } /* @@ -1234,16 +1248,12 @@ static vm_fault_t mon_bin_vma_fault(struct vm_fault *vmf) unsigned long offset, chunk_idx; struct page *pageptr; - mutex_lock(&rp->fetch_lock); offset = vmf->pgoff << PAGE_SHIFT; - if (offset >= rp->b_size) { - mutex_unlock(&rp->fetch_lock); + if (offset >= rp->b_size) return VM_FAULT_SIGBUS; - } chunk_idx = offset / CHUNK_SIZE; pageptr = rp->b_vec[chunk_idx].pg; get_page(pageptr); - mutex_unlock(&rp->fetch_lock); vmf->page = pageptr; return 0; }
The problem arises because our read() function grabs a lock of the circular buffer, finds something of interest, then invokes copy_to_user() straight from the buffer, which in turn takes mm->mmap_sem. In the same time, the callback mon_bin_vma_fault() is invoked under mm->mmap_sem. It attempts to take the fetch lock and deadlocks. This patch does away with protecting of our page list with any semaphores, and instead relies on the kernel not close the device while mmap is active in a process. In addition, we prohibit re-sizing of a buffer while mmap is active. This way, when (now unlocked) fault is processed, it works with the page that is intended to be mapped-in, and not some other random page. Note that this may have an ABI impact, but hopefully no legitimate program is this wrong. Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> Reported-by: syzbot+56f9673bb4cdcbeb0e92@syzkaller.appspotmail.com --- drivers/usb/mon/mon_bin.c | 32 +++++++++++++++++++++----------- 1 file changed, 21 insertions(+), 11 deletions(-)