qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [Qemu-ppc] [PATCH] ppc/e500_pci: Fix an array overflow


From: Scott Wood
Subject: Re: [Qemu-devel] [Qemu-ppc] [PATCH] ppc/e500_pci: Fix an array overflow issue
Date: Tue, 27 Sep 2011 11:52:39 -0500
User-agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.17) Gecko/20110428 Fedora/3.1.10-1.fc15 Thunderbird/3.1.10

On 09/27/2011 07:45 AM, Alexander Graf wrote:
> On 27.09.2011, at 10:17, Liu Yu wrote:
>> ---
>> hw/ppce500_pci.c |   26 ++++++++++++++++----------
>> 1 files changed, 16 insertions(+), 10 deletions(-)
>>
>> diff --git a/hw/ppce500_pci.c b/hw/ppce500_pci.c
>> index 2db365d..3e24e85 100644
>> --- a/hw/ppce500_pci.c
>> +++ b/hw/ppce500_pci.c
>> @@ -108,15 +108,18 @@ static uint32_t pci_reg_read4(void *opaque, 
>> target_phys_addr_t addr)
>>
>>     case PPCE500_PCI_IW3:
>>     case PPCE500_PCI_IW2:
>> -    case PPCE500_PCI_IW1:
>> +    case PPCE500_PCI_IW1: {
>> +        int idx = ((addr >> 5) & 0x3) - 1;
> 
> So this is the main change, right? Why the -1? A guest could potentially 
> access pib[-1] using this, no?

Not with the values of addr that lead to this code.  The -1 is because
IW1/2/3 are 0x1e0/0x1c0/0x1a0.  Previously IW1 would overflow the array.

>>         switch (addr & 0xC) {
>> -        case PCI_PITAR: value = pci->pib[(addr >> 5) & 0x3].pitar; break;
>> -        case PCI_PIWBAR: value = pci->pib[(addr >> 5) & 0x3].piwbar; break;
>> -        case PCI_PIWBEAR: value = pci->pib[(addr >> 5) & 0x3].piwbear; 
>> break;
>> -        case PCI_PIWAR: value = pci->pib[(addr >> 5) & 0x3].piwar; break;
>> +        case PCI_PITAR: value = pci->pib[idx].pitar; break;
>> +        case PCI_PIWBAR: value = pci->pib[idx].piwbar; break;
>> +        case PCI_PIWBEAR: value = pci->pib[idx].piwbear; break;
>> +        case PCI_PIWAR: value = pci->pib[idx].piwar; break;
> 
> I'm fairly sure this breaks checkpatch.pl.

So does the original code...

If this is to be fixed, the outbound window switch should be fixed too
(and made to use idx, for consistency).

-Scott




reply via email to

[Prev in Thread] Current Thread [Next in Thread]