Hi! I tried to write a 'Smatch' ( http://smatch.sourceforge.net/ ) script to help searching for the problem below. The question that the script would check is that is any 'struct net_device'.'name' accessed by any read-out expression before calling register_netdev(). It seems to me that 'Smatch' does not provide enough information for this: ---------------------------------------------------------------- linux-2.6.6/drivers/net/smc-ultra.c:243 is: printk("%s: %s at %#3x,", dev->name, model_name, ioaddr); smc-ultra.c.sm contains: SM drivers/net/smc-ultra.c 243 expr_stmt call_expr((addr_expr function_decl(printk))(tree_list: addr_expr string_cst(""%s: %s at %#3x,""), non_lvalue_expr parm_decl(dev), var_decl(model_name), parm_decl(ioaddr))) ---------------------------------------------------------------- and the pointer type of 'dev' is also not known: ---------------------------------------------------------------- linux-2.6.6/drivers/net/smc-ultra.c:206 is: static int __init ultra_probe1(struct net_device *dev, int ioaddr) smc-ultra.c.sm: SM drivers/net/smc-ultra.c 207 function_decl integer_type function_decl(ultra_probe1)(pointer_type(parm_decl(dev)), integer_type(parm_decl(ioaddr))) ---------------------------------------------------------------- Is it possible to update 'Smatch' to let the check possible? NMarci On Sat, 29 May 2004, Roger Luethi wrote: > On Fri, 28 May 2004 10:27:30 +0200, Németh Márton wrote: > > When I'm loading smc-ultra module with 'modprobe -k smc_ultra' using > > Linux 2.6.6 I get the following message in 'dmesg': > > > > smc-ultra.c: Presently autoprobing (not recommended) for a single card. > > smc-ultra.c: No ISAPnP cards found, trying standard ones... > > smc-ultra.c:v2.02 2/3/98 Donald Becker (becker@cesdis.gsfc.nasa.gov) > > eth%d: SMC Ultra at 0x280, 00 00 C0 78 C5 B6, IRQ 3 memory 0xd0000-0xd3fff. > > > > Using Linux kernel 2.4.25 the last line was correct, like this: > > eth0: SMC Ultra at 0x280, 00 00 C0 78 C5 B6, IRQ 3 memory 0xd0000-0xd3fff. > > > > I think the printk() in linux-2.6.6/drivers/net/smc-ultra.c:243 is buggy, > > but I do not have the knowledge how to correct this. > > There are several drivers in 2.6 that use dev->name before calling > register_netdev(), which results in the "eth%d" messages you are seeing. > I've noticed this with mace.c as well (2.6.5 IIRC). - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
