ihex: Add support for CS:IP/EIP records
PATCH 1/1 ihex: Add support for CS:IP/EIP records
ihex firmwares can include a jump address for starting execution. Add a -j option which will cause this to be written into the generated file as a record with address zero and data consisting of the address to jump to, allowing drivers to make use of this information.
This format is chosen because it most closely follows the original ihex format, though it may make more sense to write a record with length zero and the address stored as the address.
--- firmware/ihex2fw.c | 16 +++++++++++++--- 1 files changed, 13 insertions(+), 3 deletions(-)
firmware/ihex2fw.c +13 -3
@@ -55,6 +55,7 @@ static int output_records(int outfd);static int sort_records =0 ;static int wide_records =0 ; +static int include_jump =0 ;static int usage (void ) {
@@ -63,6 +64,7 @@ static int usage(void)fprintf (stderr,"usage: ihex2fw [<options>] <src.HEX> <dst.fw>\n" );fprintf (stderr," -w: wide records (16-bit length)\n" );fprintf (stderr," -s: sort records by address\n" ); +fprintf (stderr," -j: include records for CS:IP/EIP address\n" );return 1 ; }
@@ -73,7 +75,7 @@ int main(int argc, char **argv)uint8_t *data;int opt; -while ((opt =getopt (argc, argv,"ws" )) !=-1 ) { +while ((opt =getopt (argc, argv,"wsj" )) !=-1 ) {switch (opt) {case 'w' : wide_records =1 ;
@@ -81,7 +83,9 @@ int main(int argc, char **argv)case 's' : sort_records =1 ;break ; -default : +case 'j' : + include_jump =1 ; +break ;return usage (); } }
@@ -128,6 +132,7 @@ static int process_ihex(uint8_t *data, ssize_t size) {struct ihex_binrec *record;uint32_t offset =0 ; +uint32_t *data32;uint8_t type, crc =0 , crcbyte =0 ;int i, j;int line =1 ;
@@ -223,8 +228,13 @@ next_record:return -EINVAL ; } + data32 = &record->data[0 ]; + *data32 =htonl (*data32); +/* These records contain the CS/IP or EIP where execution -* starts. Don't really know what to do with them. */ +* starts. If requested output this as a record. */ +if (include_jump) +file_record (record); goto next_record;default :
Signature · 1 line
1.7.1
Quoted 4 lines · Mark Brown
ihex firmwares can include a jump address for starting execution. Add a -j option which will cause this to be written into the generated file as a record with address zero and data consisting of the address to jump to, allowing drivers to make use of this information.
Some questions...
- Which firmware blobs that are *already* in the kernel would need this? Remember, we're not adding new crap to the kernel source tree; new firmware images get added to linux-firmware.git instead.
- Why make the include_jump code optional? When would we ever *not* want to include such a record -- when does the input ihex file contain such a record that the kernel driver *must* not see because it can't cope with it?
Signature · 2 lines
David Woodhouse Open Source Technology Centre [email protected] Intel Corporation
Quoted 3 lines · David Woodhouse
- Which firmware blobs that are *already* in the kernel would need this? Remember, we're not adding new crap to the kernel source tree; new firmware images get added to linux-firmware.git instead.
None to my knowledge, however I don't understand why this would be relevant to adding new features to ihex2fw?
Quoted 4 lines
- Why make the
include_jumpcode optional? When would we ever *not* want to include such a record -- when does the input ihex file contain such a record that the kernel driver *must* not see because it can't cope with it?
Given that our binary ihex format discards all record type information I would expect all existing users would be confused if they started seeing these records so it seems safer to only generate them when required. It seems likely that some ihex images will contain jump addresses generated by their toolchain which do not need to be parsed since the address is always the same (eg, the address the device uses when brought out of reset), though none of those in the kernel tree currently seem to.
firmware/ihex2fw.c next_record: · line 232
Quoted 2 lines · On Thu, 9 Sep 2010 14:40:26 +0100 Mark Brown…
firmware/ihex2fw.c +2
-0+ data32 = &record->data[0 ]; + *data32 =htonl (*data32);
firmware/ihex2fw.c: In function 'process_ihex':firmware/ihex2fw.c:231: warning: assignment from incompatible pointer type
And afacit data[] isn't 32-bit aligned and I don't know that all architectures support htonl() against a misaligned pointer.
I'll drop the patch.
Quoted 7 lines · Andrew Morton
firmware/ihex2fw.c next_record: · line 232
Quoted 2 lines · Mark Brown
firmware/ihex2fw.c +2
-0+ data32 = &record->data[0 ]; + *data32 =htonl (*data32);
firmware/ihex2fw.c: In function 'process_ihex':firmware/ihex2fw.c:231: warning: assignment from incompatible pointer type
Which architecture are you using - my AMD64 system isn't noticing these things for me?
Quoted 2 lines
And afacit data[] isn't 32-bit aligned and I don't know that all architectures support
htonl()against a misaligned pointer.
We're pretty much safe in userspace, if the hardware doesn't support it then as far as I'm aware pretty much all OSs do a software fixup. This isn't great for performance but works.
Quoted 1 line
I'll drop the patch.
OK, I'll spin it later today.