Re: [PATCH 3/4] ArmVirtPkg/FdtClient: add methods to iterate over memory nodes

[Ard Biesheuvel <ard.biesheuvel@linaro.org>]

On 15 September 2016 at 14:57, Laszlo Ersek <lersek@redhat.com> wrote:
> On 09/15/16 15:30, Ard Biesheuvel wrote:
>> Add high level methods to iterate over all 'reg' properties of all DT
>> nodes whose device_type properties have the value "memory". Since we are
>> modifying the FdtClient protocol, update the protocol and the only existing
>> implementation at the same time.
>>
>> Contributed-under: TianoCore Contribution Agreement 1.0
>> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>> ---
>>  ArmVirtPkg/FdtClientDxe/FdtClientDxe.c  | 76 ++++++++++++++++++++
>>  ArmVirtPkg/Include/Protocol/FdtClient.h | 26 +++++++
>>  2 files changed, 102 insertions(+)
>>
>> diff --git a/ArmVirtPkg/FdtClientDxe/FdtClientDxe.c b/ArmVirtPkg/FdtClientDxe/FdtClientDxe.c
>> index 382e9af6bf84..099cce7821d6 100644
>> --- a/ArmVirtPkg/FdtClientDxe/FdtClientDxe.c
>> +++ b/ArmVirtPkg/FdtClientDxe/FdtClientDxe.c
>> @@ -194,6 +194,80 @@ FindCompatibleNodeReg (
>>
>>  STATIC
>>  EFI_STATUS
>> +EFIAPI
>> +FindNextMemoryNodeReg (
>> +  IN  FDT_CLIENT_PROTOCOL     *This,
>> +  IN  INT32                   PrevNode,
>> +  OUT INT32                   *Node,
>> +  OUT CONST VOID              **Reg,
>> +  OUT UINTN                   *AddressCells,
>> +  OUT UINTN                   *SizeCells,
>> +  OUT UINT32                  *RegSize
>> +  )
>> +{
>> +  INT32          Prev, Next;
>> +  CONST CHAR8    *DeviceType;
>> +  INT32          Len;
>> +  EFI_STATUS     Status;
>> +
>> +  ASSERT (mDeviceTreeBase != NULL);
>> +  ASSERT (Node != NULL);
>> +
>> +  for (Prev = PrevNode;; Prev = Next) {
>> +    Next = fdt_next_node (mDeviceTreeBase, Prev, NULL);
>> +    if (Next < 0) {
>> +      break;
>> +    }
>> +
>> +    DeviceType = fdt_getprop (mDeviceTreeBase, Next, "device_type", &Len);
>> +    if (DeviceType != NULL && AsciiStrCmp (DeviceType, "memory") == 0) {
>
> HighMemDxe uses AsciiStrnCmp (Type, "memory", Len) here.
>
> If we're sure that we're not looking "memory*" types here, then the
> change is fine. Are we sure?
>

Actually, that is a bug. AsciiStrCmp () is guaranteed to check the NUL
terminator, and so it will only match on "memory\0". Using
AsciiStrnCmp() with the length of the property value is guaranteed
*not* to check the NUL terminator, so it may match on prefixes of
"memory\0"




>> +
>
> The empty line is likely unintended.
>

indeed.

>> +      //
>> +      // Get the 'reg' property of this memory node. For now, we will assume
>> +      // 8 byte quantities for base and size, respectively.
>> +      // TODO use #cells root properties instead
>> +      //
>> +      Status = GetNodeProperty (This, Next, "reg", Reg, RegSize);
>> +      if (EFI_ERROR (Status)) {
>> +        DEBUG ((EFI_D_WARN,
>> +          "%a: ignoring memory node with no 'reg' property\n",
>> +          __FUNCTION__));
>> +        continue;
>> +      }
>> +      if ((*RegSize % 16) != 0) {
>> +        DEBUG ((EFI_D_WARN,
>> +          "%a: ignoring memory node with invalid 'reg' property (size == 0x%x)\n",
>> +          __FUNCTION__, RegSize));
>
> This should be *RegSize.
>

OK

> With the above confirmed / fixed:
>
> Reviewed-by: Laszlo Ersek <lersek@redhat.com>