[U-Boot-Users] reading ethaddr from env

Hello all,
I'm porting u-boot to a custom board with pxa255. I have a problem that I'm not be able to solve.
When u-boot starts and try to read the ethaddr, (lib_arm/board.c at 296), the parameter tmp used in the function getenv_r is pointing to the gd->bd structure, so the execution of this function corrupts the data stored in it.
I've found a "dirty" workaround of this problem rising the size of tmp buffer: The code in lib_arm/board.c
/* IP Address */ gd->bd->bi_ip_addr = getenv_IPaddr ("ipaddr");
/* MAC Address */ { int i; ulong reg; char *s, *e; /* uchar tmp[64]; */ uchar tmp[128];
i = getenv_r ("ethaddr", tmp, sizeof (tmp)); printf("ethaddr %s[%d]\n", tmp, i); s = (i > 0) ? tmp : NULL;
for (reg = 0; reg < 6; ++reg) { gd->bd->bi_enetaddr[reg] = s ? simple_strtoul (s, &e, 16) : 0; if (s) s = (*e) ? e + 1 : e; } }
The output of gdb before execute the getenv_r(with tmp[64]):
(gdb) print *gd $2 = {bd = 0xa00bffa0, flags = 0, baudrate = 115200, have_console = 1, reloc_off = 0, env_addr = 2685141004, env_valid = 1, fb_base = 0, jt = 0x0} (gdb) n (gdb) print &tmp[0] $3 = (uchar *) 0xa00bffa4 "" (gdb) s getenv_r (name=0xa0114bb0 "ethaddr", buf=0xa00bffa4 "À?\002?", len=64) at cmd_nvedit.c:512
As you can see the tmp buffer is pointing to the gd->bd plus 4.
My compiler:
arm-linux-gcc -v Reading specs from /home/jgarcia/PXA2XX/ELDK/eldk_work/usr/bin/../lib/gcc-lib/arm-linux/3.3.3/specs Configured with: ../configure --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --enable-shared --enable-threads=posix --disable-checking --with-system-zlib --enable-__cxa_atexit --with-newlib --enable-languages=c,c++ --disable-libgcj --host=i386-redhat-linux --target=arm-linux Thread model: posix gcc version 3.3.3 (DENX ELDK 3.1 3.3.3-8)
Any hints ?
Regards, Juan Antonio

In message 20050223100142.GA3003@nabla101 you wrote:
When u-boot starts and try to read the ethaddr, (lib_arm/board.c at 296), the parameter tmp used in the function getenv_r is pointing to the gd->bd structure, so the execution of this function corrupts the data stored in it.
You are wrong. "tmp" is a local array, declared in line 294:
294 uchar tmp[64]; 295 296 i = getenv_r ("ethaddr", tmp, sizeof (tmp));
I've found a "dirty" workaround of this problem rising the size of tmp buffer:
What exactly is your problem?
/* uchar tmp[64]; */ uchar tmp[128]; i = getenv_r ("ethaddr", tmp, sizeof (tmp));
Why do you think that 64 bytes should be insufficient to store the MAC address?
As you can see the tmp buffer is pointing to the gd->bd plus 4.
At least this is what the debugger claims.
Any hints ?
I cannot reproduce this problem here. WHat do you get when you use printf() to display the addresses _before_ calling getenv_r()?
Best regards,
Wolfgang Denk

On 24/02/05 23:49, Wolfgang Denk wrote:
In message 20050223100142.GA3003@nabla101 you wrote:
I cannot reproduce this problem here. WHat do you get when you use printf() to display the addresses _before_ calling getenv_r()?
I've found the problem. I'd forgotten to assign the CFG_GBL_DATA_SIZE constant, so sp was pointing to a wrong place.
Thanks for your answer.
Regards, Juan Antonio
participants (2)
-
Juan Antonio Garcia Redondo
-
Wolfgang Denk