Reverted 'mfr_id' to 'manufacturer_id'
Moved devices table definition to flashmem.h
Single global variable 'spi_flash_pages64k'
The local structure holding the actual device data is used in firmware.c only
difference in code:
```
text data bss dec hex filename
223189 138560 6067 367816 59cc8 ./armsrc/obj/fullimage.elf <-- c9e751d darkside: fix backdoor support
223437 138624 6075 368136 59e08 ./armsrc/obj/fullimage.elf <-- this commit
======================================
+248 +64 +8 +320
```
* introduced a new communication command CMD_FLASHMEM_PAGES64K to get the number of 64k pages
* "the last page" is a special, holding the dicts and some other stuff, relocated to there
* raised timeout when wiping a mem page (W25Q16 looks a bit slower than W25X20BV)
* loop all pages in Flash_WipeMemory()
this could happen if TC2 was already 0 when it was started or reset
resulting in the initial reset not happening until TC0 had overflowed for
the first time as the delay to ensure this didn't happen would be missed
when TC2 was already 0
the new behaviour results in TIOA0 being toggled when a software trigger
of TC0 happens which makes TC2 reset immediately without having to wait
for TC0 to overflow
Now, you can enable at least two of your favorite technologies (such as LF and HF 14443A) attached a standalone mode and still have spare ROM space for other functionalities on a Proxmark3 Easy with a 256KiB ROM.
Check if memory allocation fails
Fix memory leak
Initialize struct in declaration
Add/Fix some notes
Remove unlikely() in favor of readability
Remove a hard-coded magic number
Before:
708608,708608,707584,707584,708608->708198.4 bytes/s
After:
722944,733184,732160,731136,733184->730521.6 bytes/s
Added some note for ping-pong mode and non ping-pong mode
To make the full use of the ping-pong endpoint
Transfer speed before this fix:
616448,618496,618496,615424,615424->616857.6 bytes/s
Transfer speed after this fix:
707584,709632,707584,709632,710656->709017.6 bytes/s (+14.9%)
Tested by running hw status