On 2023-11-15 18:27, Mike Maslenkin wrote:
On Wed, Nov 15, 2023 at 4:52 PM Igor Kulchytskyy <ig...@ami.com> wrote:
Hello Leif and Mike,
Let me try to explain the idea of the filtering IP.
That filtering should work only if we know exactly that IP is IPv4 or IPv6 in
SMBIOS Type 42.
Hm. I've already composed a reply below, but then a returned to this
statement...
Is this a difference in condition between v3 and v5? I came to the
conclusion that at the place we are discussing
SMBIOS table 42h can be absent because
PlatformHostInterfaceInformationReady hasn't been called yet,
so REDFISH_HOST_INTERFACE_HOST_IP_ADDRESS_FORMAT_UNKNOWN is expected.
And it just helping to reduce the work in case we know the exact type of IP,
which supposed to be used in BIOS BMC communication.
In that case there is no need to build network interface for the unused IP Type.
On some systems IP address could be dynamic and we will not be able to know the
version of IP from SMBIOS.
If you see I check HostIpAssignmentType in GetHiIpProtocolType function. And
return IP type UNKNOWN if it is not static.
If we get an unknown IP type, we should not return from BuildupNetworkInterface
function, but just proceed and build the network interface for all IPs.
So, later Redfish Discover driver can find the right one.
Based on this logic I'm going to prepare the patch v6.
Thank you,
Igor
Agree.. I was focused on edk2 implementation of
RedfishPlatformHostInterfaceLib and PlatformHostInterfaceBmcUsbNicLib
where HostIpAddressFormat is specified (hardcoded). I guess
HostIpAddressFormat as well as SMBIOS table 42h must be available
by the time RedfishServiceAcquireService()->DiscoverRedfishHostInterface()
call, while it might be not available at the moment
RedfishDiscoverDriverBindingStart()->BuildupNetworkInterface(). So,
condition from v3 looks correct to me.
My main concern was introduction of defines. Those don't look great.
Those are huge (it even doesn't fit into the screen) and misleading a
bit.
For example:
+#define MAC_COMPARE(ThisNetworkInterface, TargetNetworkInterface)
(CompareMem ((VOID *)&ThisNetworkInterface->MacAddress,
&TargetNetworkInterface->MacAddress,
ThisNetworkInterface->HwAddressSize))
The proposed variant is equal to #define MAC_COMPARE(A, B)
(CompareMem ((VOID *)&ThisNetworkInterface->MacAddress,
&TargetNetworkInterface->MacAddress,
ThisNetworkInterface->HwAddressSize)), i.e a bit useless.
I would expect it could be declared at least as:
#define MAC_COMPARE(This, Target) CompareMem ((VOID
*)&(This)->MacAddress, &(Target)->MacAddress, (This)->HwAddressSize)
I.e define should really replace some arguments also reducing the line length.
BTW: there is a place in ValidateTargetNetworkInterface() where
CompareMem can be replaced with MAC_COMPARE too.
Also, I found IP6_LINK_EQUAL(Mac1, Mac2) define, that is unused in
edk2. But according to that one, please consider moving "== 0" check
to #define declaration.
But I do not think this macro is required at all, because there are 5
MAC compares left in this module. So, it just brings some
inconsistency.
Agreed that static helper function would be the best.
Leif, do you expect something like this?
STATIC
BOOLEAN
FilterInterface (
IN NETWORK_INTERFACE_PROTOCOL_TYPE ProtocolType,
IN UINT8 HostIpAddressFormat
)
{
// This is based on v5, but according to the comments above
// v3 is correct as it allows
REDFISH_HOST_INTERFACE_HOST_IP_ADDRESS_FORMAT_UNKNOWN
if (ProtocolType == ProtocolTypeTcp4) {
return IpType != REDFISH_HOST_INTERFACE_HOST_IP_ADDRESS_FORMAT_IP4;
} else if (ProtocolType == ProtocolTypeTcp6) {
return IpType != REDFISH_HOST_INTERFACE_HOST_IP_ADDRESS_FORMAT_IP6;
}
Yes, this looks ideal.
return false;
Although this should be FALSE (upper-case).
}
and then::
// Get IP Type to filter out unnecessary network protocol if possible
IpType = GetHiIpProtocolType ();
for (Index = 0; Index < ListCount; Index++) {
// Check IP Type and skip an unnecessary network protocol if does not match
if (FilterInterface (gRequiredProtocol[Index].ProtocolType, IpType)) {
This gives us a big but still readable chunk here, and much neater tests
in the helper than they *could* be either in place or in macros.
continue;
}
Ship it.
/
Leif
Regards,
Mike.
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#111314): https://edk2.groups.io/g/devel/message/111314
Mute This Topic: https://groups.io/mt/102584140/21656
Group Owner: devel+ow...@edk2.groups.io
Unsubscribe:
https://edk2.groups.io/g/devel/leave/9847357/21656/1706620634/xyzzy
[arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-
