Re: [edk2-devel] [PATCH v1] NetworkPkg/UefiPxeBcDxe: Fix PXE_BOOT_SERVERS usage in boot info parse flow

["Laszlo Ersek" <lersek@redhat.com>]

On 08/19/20 20:13, Laszlo Ersek wrote:
> On 08/19/20 18:53, Maciej Rabeda wrote:
>> REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2876
>>
>> According to PXE 2.1 spec, DHCP option 43, tag 6 (PXE_DISCOVERY_CONTROL),
>> bit 3 specifies whether PXE client should use/accept TFTP servers defined
>> within PXE_BOOT_SERVERS list (tag 8). This bit was not being taken into
>> account when choosing boot server IP in PxeBcDhcp4BootInfo().
>>
>> Cc: Jiaxin Wu <jiaxin.wu@intel.com>
>> Cc: Siyuan Fu <siyuan.fu@intel.com>
>> Signed-off-by: Maciej Rabeda <maciej.rabeda@linux.intel.com>
>> ---
>>  NetworkPkg/UefiPxeBcDxe/PxeBcBoot.c | 10 +++++++---
>>  1 file changed, 7 insertions(+), 3 deletions(-)
>>
>> diff --git a/NetworkPkg/UefiPxeBcDxe/PxeBcBoot.c b/NetworkPkg/UefiPxeBcDxe/PxeBcBoot.c
>> index d062a526077b..ed9bca0f7de3 100644
>> --- a/NetworkPkg/UefiPxeBcDxe/PxeBcBoot.c
>> +++ b/NetworkPkg/UefiPxeBcDxe/PxeBcBoot.c
>> @@ -499,12 +499,16 @@ PxeBcDhcp4BootInfo (
>>  
>>    //
>>    // Parse the boot server address.
>> -  // If prompt/discover is disabled, get the first boot server from the boot servers list.
>> -  // Otherwise, parse the boot server Ipv4 address from next server address field in DHCP header.
>> +  // If prompt/discover is disabled, server list should be used and is present (DHCP option 43),
>> +  // get the first boot server from the boot servers list.
>> +  // Otherwise, parse the boot server IPv4 address from next server address field in DHCP header.
>>    // If all these fields are not available, use option 54 instead.
>>    //
>>    VendorOpt = &Cache4->VendorOpt;
>> -  if (IS_DISABLE_PROMPT_MENU (VendorOpt->DiscoverCtrl) && IS_VALID_BOOT_SERVERS (VendorOpt->BitMap)) {
>> +  if (IS_DISABLE_PROMPT_MENU (VendorOpt->DiscoverCtrl) &&
>> +      IS_VALID_BOOT_SERVERS (VendorOpt->BitMap) &&
>> +      IS_ENABLE_USE_SERVER_LIST (VendorOpt->DiscoverCtrl))
>> +  {
>>      Entry = VendorOpt->BootSvr;
>>      if (VendorOpt->BootSvrLen >= sizeof (PXEBC_BOOT_SVR_ENTRY) && Entry->IpCnt > 0) {
>>        CopyMem (
>>
> 
> I'm still undecided whether option#43 / tag#6 / bit#3 being clear means
> we should *ignore* PXE_BOOT_SERVERS (tag#8), but I'm willing to defer to
> you on that. So, I can give a cautious
> 
> Reviewed-by: Laszlo Ersek <lersek@redhat.com>
> 
> for this patch.
> 
> Regarding the feature freeze -- in theory, this is a bugfix. However,
> before we merge it, it would be really nice to get feedback from the
> original reporter (CC'd).
> 
> I also intend to regression-test this patch, I'll report back.

I've repeated my usual PXEv4 boot tests (using OVMF IA32X64 and
ArmVirtQemu AARCH64 guests, covering shim + grub + vmlinuz + initrd);
nothing seems to have regressed.

Regression-tested-by: Laszlo Ersek <lersek@redhat.com>

Thanks
Laszlo