Endianness assumptions

The conversions.h: these macros assume the cpu is little endian, that is wrong for ARM

replace INT32_TO_BUFFER(a,b) as call to *(int32_t*)b=htonl(a); maybe better
furthermore, in the same file hton_int32() is not necessary: the htonl() libc library do the same

two side points:

If we consider (n)FAPI, the problem is a bit different: FAPI can be "native" because it runs on same machine, this is another possible choice: native against nFAPI order (probably big endian, but i have not verified)

Performance: if performance matter, these byte swaps are available in SIMD, for a performance x10
openair3/NAS/UE/ESM/esm_ebr_context.c:215 => this one, no doubts, htonl/ntohl should be used the developer is wrong
openair2/UTIL/OSA/osa_internal.h:43: yet another inet_to32 surprisingly not used in the same block of code: openair2/UTIL/OSA/osa_snow3g.c:144: return ( ( ((uint32_t)r0) << 24 ) | ( ((uint32_t)r1) << 16 ) | ( ((uint32_t)r2) << 8 ) | ( ((uint32_t)r3) ) ); for example
i still missed to removed some duplicated code: openair3/SECU/snow3g.c => same "error" in duplicated file
openair3/S1AP/s1ap_eNB.c:121 also clearly assuming endianness
nfapi/open-nFAPI/nfapi/src/nfapi.c:224 same
why is it a macro ? a C function would be normal C code and allows to verify parameters. i don't like the name: maybe encode_eNBid_to_asn1c()? the calloc should of course be malloc (erases all values just after)

#define 
MACRO_ENB_ID_TO_BIT_STRING(mACRO, bITsTRING)    \
do {                                                    \
    (bITsTRING)->buf = calloc(3, sizeof(uint8_t));      \
    (bITsTRING)->buf[0] = ((mACRO) >> 12);              \
    (bITsTRING)->buf[1] = (mACRO) >> 4;                 \
    (bITsTRING)->buf[2] = ((mACRO) & 0x0f) << 4;        \
    (bITsTRING)->size = 3;                              \
    (bITsTRING)->bits_unused = 4;                       \
} while(0)

0 of 9 checklist items completed · Edited 2 years ago

Designs

Child items 0

No child items are currently assigned. Use child items to break down this issue into smaller parts.

Endianness assumptions

Activity