Re: [PATCH v2 07/24] EDAC: Kill EDAC_DIMM_OFF() macro

[Borislav Petkov]

On Mon, Jun 24, 2019 at 03:09:09PM +0000, Robert Richter wrote:
> We do not need to calculate the offset in the mc's dimm array, let's
> just store the index in struct dimm_info and we can get rid of this
> macro.
> 
> Signed-off-by: Robert Richter <rrichter@xxxxxxxxxxx>
> ---
>  drivers/edac/edac_mc.c       | 13 ++++--------
>  drivers/edac/edac_mc_sysfs.c | 20 ++++--------------
>  include/linux/edac.h         | 41 ------------------------------------
>  3 files changed, 8 insertions(+), 66 deletions(-)

I like this cleanup a lot. Good!

> diff --git a/drivers/edac/edac_mc.c b/drivers/edac/edac_mc.c
> index c959e8b1643c..c44bc83e8502 100644
> --- a/drivers/edac/edac_mc.c
> +++ b/drivers/edac/edac_mc.c
> @@ -318,7 +318,7 @@ struct mem_ctl_info *edac_mc_alloc(unsigned mc_num,
>  	unsigned size, tot_dimms = 1, count = 1;
>  	unsigned tot_csrows = 1, tot_channels = 1, tot_errcount = 0;
>  	void *pvt, *p, *ptr = NULL;
> -	int i, j, row, chn, n, len, off;
> +	int idx, i, j, row, chn, n, len;
>  	bool per_rank = false;
>  
>  	BUG_ON(n_layers > EDAC_MAX_LAYERS || n_layers == 0);
> @@ -426,20 +426,15 @@ struct mem_ctl_info *edac_mc_alloc(unsigned mc_num,
>  	memset(&pos, 0, sizeof(pos));
>  	row = 0;
>  	chn = 0;
> -	for (i = 0; i < tot_dimms; i++) {
> +	for (idx = 0; idx < tot_dimms; idx++) {
>  		chan = mci->csrows[row]->channels[chn];
> -		off = EDAC_DIMM_OFF(layer, n_layers, pos[0], pos[1], pos[2]);
> -		if (off < 0 || off >= tot_dimms) {
> -			edac_mc_printk(mci, KERN_ERR, "EDAC core bug: EDAC_DIMM_OFF is trying to do an illegal data access\n");
> -			goto error;
> -		}

Btw, right around here there's a comment:

        /*
         * Allocate and fill the dimm structs
         */

and since you're cleaning up all this, can you please add another patch
which takes all that code under the comment and see if you can carve it
out into a separate helper edac_alloc_dimms() or so. Because that
edac_mc_alloc() function is huuuge.

Btw, the code under

        /*
         * Alocate and fill the csrow/channels structs
         */

begs to be a separate function too :-)

Thx.

-- 
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.