Re: Linux-2.6.13-rc7

[Al Viro <viro@parcelfarce.linux.theplanet.co.uk>]

On Thu, Aug 25, 2005 at 11:27:32AM +0400, Alexey Dobriyan wrote:
> Mine is alpha-unknown-linux-gnu-gcc (GCC) 3.4.4 (Gentoo 3.4.4)
> 
> > Which place triggers it in your build?
> 
> net/ipv4/route.c:3152, call to rt_hash_lock_init().
> 
> >From preprocessed source (reformatted):
> -----------------------------------------------------------------------
> typedef struct {
> 	volatile unsigned int lock;
> 
> 	int on_cpu;
> 	int line_no;
> 	void *previous;
> 	struct task_struct * task;
> 	const char *base_file;
> } spinlock_t;
> 
> static inline void *kmalloc(size_t size, unsigned int flags)

Oh, lovely...

a) gcc4 on alpha refuses to make that inline
b) bug is real, indeed - spinlock debugging + >32 CPU => panic in ip_rt_init()

IMO that's a question to rth: why do we really need to block always_inline
on alpha?