SCSI_NETWORK_WIFI_CMD_SCAN_RESULTS ignores transfer size

[th-otto]

In https://github.com/BlueSCSI/BlueSCSI-v2/blob/df274ea5aae5b7b2cd57a24ffafe9c981f4490a5/lib/SCSI2SD/src/firmware/network.c#L328, the size of the returned data is only checked against the size of the internal data buffer. But that ignores the size of the initiators specified transfer size in bytes 3/4 of the CDB.

I don't know if this can cause trouble, since at most 10 wifi_network_entries are returned (using 2 + 10 * 74 bytes), but IMHO this should be fixed.

BTW, is there any way to increase that number of entries (WIFI_NETWORK_LIST_ENTRY_COUNT)?

[erichelgeson]

Agree, should truncate the response if the host asks for a specific len.

WIFI_NETWORK_LIST_ENTRY_COUNT is 10 currently and it was just a nice round number to not take up too much RAM, as we are limited on the Pico. I suppose in apartments or the like there could be a lot more than 10 around you.

If you'd like to PR the fix, we'd love to have it!

[th-otto]

I understand the limitation, but the problem is that even "at home", their may be entries from your neighbours that you are not interested in (unless you know their passwords ;)

I guess its too late to change the interface, so the burden of storing/filtering the entries can be put to the host application instead?

[erichelgeson]

Just brainstorming.. since SCSI_NETWORK_WIFI_CMD_SCAN_RESULTS is the only place that wifi_network_list is used - instead of storing it in the wifi_network_list we can write entries directly to scsiDev.data (as is done in many places.. for better or worse) to save memory. This could be done in a compatible/non-breaking change to the existing mac wifi-da.

[jcs]

I think the intention was to sort by RSSI so that closer/stronger networks would appear first in the list, making it less likely that the one you care about is far down the list. But I guess I never wrote that.

https://github.com/BlueSCSI/BlueSCSI-v2/blob/df274ea5aae5b7b2cd57a24ffafe9c981f4490a5/lib/BlueSCSI_platform_RP2040/BlueSCSI_platform_network.cpp#L173

[th-otto]

instead of storing it in the wifi_network_list we can write entries directly to scsiDev.data

But is that safe? What if some other command is sent, while the scan is in progress (possibly from some other application).

[erichelgeson]

But is that safe? What if some other command is sent, while the scan is in progress (possibly from some other application).

Edit: whoops I was mistaken, you're right, we cant do this.

[th-otto]

PR just opened. I've no way to test it, but checked that it compiles atleast.