Re: [Qemu-devel] [PATCH v2] pci: Common overflow prevention

[Isaku Yamahata]

On Thu, Jul 28, 2011 at 11:40:21AM +0300, Michael S. Tsirkin wrote:
> On Thu, Jul 28, 2011 at 04:23:24PM +0900, Isaku Yamahata wrote:
> > This might be a bit late comment...
> > 
> > On Fri, Jul 22, 2011 at 11:05:01AM +0200, Jan Kiszka wrote:
> > > diff --git a/hw/pci_host.c b/hw/pci_host.c
> > > index 728e2d4..bfdc321 100644
> > > --- a/hw/pci_host.c
> > > +++ b/hw/pci_host.c
> > > @@ -47,17 +47,33 @@ static inline PCIDevice *pci_dev_find_by_addr(PCIBus 
> > > *bus, uint32_t addr)
> > >      return pci_find_device(bus, bus_num, devfn);
> > >  }
> > >  
> > > +void pci_config_write_common(PCIDevice *pci_dev, uint32_t addr,
> > > +                             uint32_t limit, uint32_t val, uint32_t len)
> > > +{
> > > +    assert(len <= 4);
> > > +    pci_dev->config_write(pci_dev, addr, val, MIN(len, limit - addr));
> > > +}
> > > +
> > > +uint32_t pci_config_read_common(PCIDevice *pci_dev, uint32_t addr,
> > > +                                uint32_t limit, uint32_t len)
> > > +{
> > > +    assert(len <= 4);
> > > +    return pci_dev->config_read(pci_dev, addr, MIN(len, limit - addr));
> > > +}
> > > +
> > 
> > Since limit and addr is unsigned, MIN(len, limit - addr) = len if limit < 
> > addr.
> > So we need explicit "if (limit < addr) return;".
> > Here's the patch for pci branch.
> > 
> > >From 75c1a2b47c93ad987cd7a37fb62bda9a59f27948 Mon Sep 17 00:00:00 2001
> > Message-Id: <address@hidden>
> > From: Isaku Yamahata <address@hidden>
> > Date: Thu, 28 Jul 2011 16:20:28 +0900
> > Subject: [PATCH] pci/host: limit check of pci_host_config_read/write_common
> > 
> > This patch adds boundary check in pci_host_config_read/write_common()
> > Since limit and addr is unsigned, MIN(len, limit - addr) = len if limit < 
> > addr.
> > So we need explicit "if (limit <= addr) return;"
> > 
> > Signed-off-by: Isaku Yamahata <address@hidden>
> 
> I don't see a problem with this, but could you please clarify when does
> this happen? I think this is only possible for a pci device
> behind an express root. If so, this belongs in pcie_host.c

Right. I'll move the check into pcie_host.c

-- 
yamahata