[ofa-general] OFED, the backported <linux/scatterlist.h> header and sg_init_table()

[Jack Morgenstein]

On Saturday 02 May 2009 14:46, Bart Van Assche wrote:
> Hello,
> 
> Yesterday I installed OFED-1.4.1-rc4 on a CentOS 5.3 system and started
> looking at the backported kernel headers. I found the following in the
> header file
> /usr/src/ofa_kernel-1.4.1/kernel_addons/backport/2.6.18-EL5.3/include/linux/scatterlist.h:
> 
> #define sg_init_table(a, b)
> 
> Or: sg_init_table() is defined to do nothing. I was expecting the following
> however:
> 
> #define sg_init_table(sgl, nents) memset(sgl, 0, sizeof(*sgl) * nents);
> 
> The sg_init_table() function is implemented in e.g. 2.6.29 as follows:
> 
> void sg_init_table(struct scatterlist *sgl, unsigned int nents)
> {
>         memset(sgl, 0, sizeof(*sgl) * nents);
> #ifdef CONFIG_DEBUG_SG
>         {
>                 unsigned int i;
>                 for (i = 0; i < nents; i++)
>                         sgl[i].sg_magic = SG_MAGIC;
>         }
> #endif
>         sg_mark_end(&sgl[nents - 1]);
> }
> 
> Does anyone know why sg_init_table() is defined such that it does nothing in
> the backported OFED headers ?
> 
> Bart.

I checked this more carefully.
Use of sg_init_table was introduced in 2.6.24 by Jens Axboe, in commit
45711f1af6eff1a6d010703b4862e0d2b9afd056. (see chunks for core/umem.c)

Before this, no initialization was done on the sg page_list, and we had no
problems.  When doing the backport, then, I simply made this a NOP.
I'm not convinced that sg_init_table needs to be implemented in kernels earlier
than 2.6.24, since this call is not replacing anything (e.g., a kzalloc), and
the page list was not previously zeroed out before usage.

What do you think?

- Jack