Attention is currently required from: laforge. jtavares has posted comments on this change. ( https://gerrit.osmocom.org/c/osmo-remsim/+/30139 )
Change subject: bankd: Add GSMTAP functionality for SIM traffic ......................................................................
Patch Set 2:
(5 comments)
Patchset:
PS2: Responses to comments. New patchset to follow in a bit.
File src/bankd/bankd_main.c:
https://gerrit.osmocom.org/c/osmo-remsim/+/30139/comment/87b7e619_e407aed0 PS2, Line 331: c = getopt_long(argc, argv, "hVd:i:p:b:n:N:I:P:sg:G:", long_options, &option_index);
Patchset 2 note: I removed bogus option 'o'. […]
Done
File src/bankd/gsmtap.c:
https://gerrit.osmocom.org/c/osmo-remsim/+/30139/comment/cb20c084_d2a3ffe0 PS2, Line 28: osmo_bankd_gsmtap_init
there's a problem with the naming prefix, sorry. […]
No problem. So, I largely copied this file from the simtrace2 project, and renamed osmo_st2_gsmtap_*() to osmo_bankd_gsmtap_*(). So, these functions are incorrectly named there, as well. The only other change I made was to allow two distinct buffers to be passed in on send_apdu() (one for the request and one for the response).
I'll get this cleaned up.
https://gerrit.osmocom.org/c/osmo-remsim/+/30139/comment/e0093a1a_451a46d1 PS2, Line 35: perror("unable to open GSMTAP");
normally we'd want to use the libosmocore logging framework with LOGL_ERROR here.
Will do. As mentioned previously, this was copied from simtrace2 project, which might not be as tightly integrated with libosmocore.
https://gerrit.osmocom.org/c/osmo-remsim/+/30139/comment/650161ce_8c70cfb9 PS2, Line 56: uint8_t *buf = malloc(gross_len);
not critical, just a comment: we could do without a dynamic heap allocation + memcpy here if we were […]
Nice-- I'd never heard of writev(). Very useful. I agree that this is not likely a performance concern given the kind of hardware a bank daemon is likely run on, but I'll make it happen, at the very least to learn something new.