Fix the handling of MEM_FRAG_TX/RX SMCs when the full memory descriptor
does not fit in a single innovation of a memory sharing request.
The current implementation expects a FFA_MEM_SHARE/FFA_MEM_LEND
call to always receive a FFA_SUCCESS response, however in the
case where a full descriptor does not fit inside the partitions
TX buffer, the call can instead complete with a FFA_MEM_FRAG_RX SMC
to request the next part of the descriptor to be transmitted.
Similarly a FFA_MEM_FRAG_TX call currently only expects
FFA_MEM_FRAG_RX as a response, however once the full descriptor
has been transmitted the FFA_SUCCESS ABI will be used to indicate
successful transmission.
Update the existing code to match the expected behaviour.
Link: https://lore.kernel.org/r/20220426121219.1801601-1-marc.bonnici@arm.com
Signed-off-by: Marc Bonnici <marc.bonnici@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
if (ret.a0 == FFA_ERROR)
return ffa_to_linux_errno((int)ret.a2);
- if (ret.a0 != FFA_SUCCESS)
+ if (ret.a0 == FFA_SUCCESS) {
+ if (handle)
+ *handle = PACK_HANDLE(ret.a2, ret.a3);
+ } else if (ret.a0 == FFA_MEM_FRAG_RX) {
+ if (handle)
+ *handle = PACK_HANDLE(ret.a1, ret.a2);
+ } else {
return -EOPNOTSUPP;
-
- if (handle)
- *handle = PACK_HANDLE(ret.a2, ret.a3);
+ }
return frag_len;
}
if (ret.a0 == FFA_ERROR)
return ffa_to_linux_errno((int)ret.a2);
- if (ret.a0 != FFA_MEM_FRAG_RX)
- return -EOPNOTSUPP;
+ if (ret.a0 == FFA_MEM_FRAG_RX)
+ return ret.a3;
+ else if (ret.a0 == FFA_SUCCESS)
+ return 0;
- return ret.a3;
+ return -EOPNOTSUPP;
}
static int