[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250910150525.GE11602@yaz-khff2.amd.com>
Date: Wed, 10 Sep 2025 11:05:25 -0400
From: Yazen Ghannam <yazen.ghannam@....com>
To: Avadhut Naik <avadhut.naik@....com>
Cc: linux-edac@...r.kernel.org, bp@...en8.de, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 2/5] EDAC/amd64: Remove NUM_CONTROLLERS macro
On Tue, Sep 09, 2025 at 06:53:11PM +0000, Avadhut Naik wrote:
> Currently, the NUM_CONTROLLERS macro is only used to statically allocate
> the csels array of struct chip_select in struct amd64_pvt.
>
> The size of this array, however, will never exceed the number of UMCs on
> the SOC. Since, max_mcs variable in struct amd64_pvt already stores the
> number of UMCs on the SOC, the macro can be removed and the static array
> can be dynamically allocated instead.
You should note that max_mcs and the csels array are also used in legacy
systems with 'DCTs'.
Those had a max of 2 controllers which we already set in
per_family_init() as the global default. So the legacy systems are
covered by this change too.
Without noting this, it seems like that case may be overlooked.
>
> Signed-off-by: Avadhut Naik <avadhut.naik@....com>
> ---
> Changes in v3:
> Patch introduced.
> ---
> drivers/edac/amd64_edac.c | 19 +++++++++++++------
> drivers/edac/amd64_edac.h | 5 ++---
> 2 files changed, 15 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c
> index 3989794e4f29..0fade110c3fb 100644
> --- a/drivers/edac/amd64_edac.c
> +++ b/drivers/edac/amd64_edac.c
> @@ -4000,30 +4000,34 @@ static int probe_one_instance(unsigned int nid)
> if (ret < 0)
> goto err_enable;
>
> + pvt->csels = kcalloc(pvt->max_mcs, sizeof(*pvt->csels), GFP_KERNEL);
> + if (!pvt->csels)
> + goto err_enable;
> +
You can move this allocation to the end of per_family_init(). That's
where we determine 'max_mcs'.
If you do so, then the 'goto' changes below are not needed.
Another option is to put it in hw_info_get() like we do for UMCs. But
that means adding the allocation to three different helper functions
rather than just the one with per_family_init().
> ret = pvt->ops->hw_info_get(pvt);
> if (ret < 0)
> - goto err_enable;
> + goto err_csels;
>
> ret = 0;
> if (!instance_has_memory(pvt)) {
> amd64_info("Node %d: No DIMMs detected.\n", nid);
> - goto err_enable;
> + goto err_csels;
> }
>
> if (!pvt->ops->ecc_enabled(pvt)) {
> ret = -ENODEV;
>
> if (!ecc_enable_override)
> - goto err_enable;
> + goto err_csels;
>
> if (boot_cpu_data.x86 >= 0x17) {
> amd64_warn("Forcing ECC on is not recommended on newer systems. Please enable ECC in BIOS.");
> - goto err_enable;
> + goto err_csels;
> } else
> amd64_warn("Forcing ECC on!\n");
>
> if (!enable_ecc_error_reporting(s, nid, F3))
> - goto err_enable;
> + goto err_csels;
> }
>
> ret = init_one_instance(pvt);
> @@ -4033,7 +4037,7 @@ static int probe_one_instance(unsigned int nid)
> if (boot_cpu_data.x86 < 0x17)
> restore_ecc_error_reporting(s, nid, F3);
>
> - goto err_enable;
> + goto err_csels;
> }
>
> amd64_info("%s detected (node %d).\n", pvt->ctl_name, pvt->mc_node_id);
> @@ -4043,6 +4047,8 @@ static int probe_one_instance(unsigned int nid)
>
> return ret;
>
> +err_csels:
> + kfree(pvt->csels);
This can go in hw_info_put(). We have kfree(pvt->umc) there already.
> err_enable:
> hw_info_put(pvt);
> kfree(pvt);
> @@ -4077,6 +4083,7 @@ static void remove_one_instance(unsigned int nid)
> /* Free the EDAC CORE resources */
> mci->pvt_info = NULL;
>
> + kfree(pvt->csels);
> hw_info_put(pvt);
> kfree(pvt);
> edac_mc_free(mci);
> diff --git a/drivers/edac/amd64_edac.h b/drivers/edac/amd64_edac.h
> index 56999ed3ae56..39d30255c767 100644
> --- a/drivers/edac/amd64_edac.h
> +++ b/drivers/edac/amd64_edac.h
> @@ -96,7 +96,6 @@
> /* Hardware limit on ChipSelect rows per MC and processors per system */
> #define NUM_CHIPSELECTS 8
> #define DRAM_RANGES 8
> -#define NUM_CONTROLLERS 12
>
> #define ON true
> #define OFF false
> @@ -347,8 +346,8 @@ struct amd64_pvt {
> u32 dbam0; /* DRAM Base Address Mapping reg for DCT0 */
> u32 dbam1; /* DRAM Base Address Mapping reg for DCT1 */
>
> - /* one for each DCT/UMC */
> - struct chip_select csels[NUM_CONTROLLERS];
> + /* Allocate one for each DCT/UMC */
> + struct chip_select *csels;
>
> /* DRAM base and limit pairs F1x[78,70,68,60,58,50,48,40] */
> struct dram_range ranges[DRAM_RANGES];
> --
Thanks,
Yazen
Powered by blists - more mailing lists