[PATCH v3 3/6] EDAC/amd64: Support more than two Unified Memory Controllers

From: Ghannam, Yazen
Date: Thu Feb 28 2019 - 10:36:47 EST


From: Yazen Ghannam <yazen.ghannam@xxxxxxx>

The first few models of Family 17h all had 2 Unified Memory Controllers
per Die, so this was treated as a fixed value. However, future systems
may have more Unified Memory Controllers per Die.

Related to this, the channel number and base address of a Unified Memory
Controller were found by matching on fixed, known values. However,
current and future systems follow this pattern for the channel number
and base address of a Unified Memory Controller: 0xYXXXXX, where Y is
the channel number. So matching on hardcoded values is not necessary.

Set the number of Unified Memory Controllers at driver init time based
on the Family/Model. Also, update the functions that find the channel
number and base address of a Unified Memory Controller to support more
than two.

Signed-off-by: Yazen Ghannam <yazen.ghannam@xxxxxxx>
---
Link:
https://lkml.kernel.org/r/20190226172532.12924-2-Yazen.Ghannam@xxxxxxx

v2->v3:
* Apply V2 Patch 3 before V2 Patch 2.

v1->v2:
* Fix tone in commit message.
* Clarify pattern used for finding channel numbers.
* Remove macro for looping over number of UMCs.
* Move function to find number of UMCs to single driver init function.
* Rename function that finds the number of UMCs.
* Add comments for new variables and functions.
* Move function to find number of UMCs out of header.

drivers/edac/amd64_edac.c | 44 ++++++++++++++++++++++++---------------
drivers/edac/amd64_edac.h | 10 ++++++---
2 files changed, 34 insertions(+), 20 deletions(-)

diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c
index e4fd459d807a..25416c608908 100644
--- a/drivers/edac/amd64_edac.c
+++ b/drivers/edac/amd64_edac.c
@@ -450,7 +450,7 @@ static void get_cs_base_and_mask(struct amd64_pvt *pvt, int csrow, u8 dct,
for (i = 0; i < pvt->csels[dct].m_cnt; i++)

#define for_each_umc(i) \
- for (i = 0; i < NUM_UMCS; i++)
+ for (i = 0; i < num_umcs; i++)

/*
* @input_addr is an InputAddr associated with the node given by mci. Return the
@@ -2476,18 +2476,14 @@ static inline void decode_bus_error(int node_id, struct mce *m)
* To find the UMC channel represented by this bank we need to match on its
* instance_id. The instance_id of a bank is held in the lower 32 bits of its
* IPID.
+ *
+ * Currently, we can derive the channel number by looking at the 6th nibble in
+ * the instance_id. For example, instance_id=0xYXXXXX where Y is the channel
+ * number.
*/
-static int find_umc_channel(struct amd64_pvt *pvt, struct mce *m)
+static int find_umc_channel(struct mce *m)
{
- u32 umc_instance_id[] = {0x50f00, 0x150f00};
- u32 instance_id = m->ipid & GENMASK(31, 0);
- int i, channel = -1;
-
- for (i = 0; i < ARRAY_SIZE(umc_instance_id); i++)
- if (umc_instance_id[i] == instance_id)
- channel = i;
-
- return channel;
+ return (m->ipid & GENMASK(31, 0)) >> 20;
}

static void decode_umc_error(int node_id, struct mce *m)
@@ -2509,11 +2505,7 @@ static void decode_umc_error(int node_id, struct mce *m)
if (m->status & MCI_STATUS_DEFERRED)
ecc_type = 3;

- err.channel = find_umc_channel(pvt, m);
- if (err.channel < 0) {
- err.err_code = ERR_CHANNEL;
- goto log_error;
- }
+ err.channel = find_umc_channel(m);

if (umc_normaddr_to_sysaddr(m->addr, pvt->mc_node_id, err.channel, &sys_addr)) {
err.err_code = ERR_NORM_ADDR;
@@ -3252,6 +3244,22 @@ static const struct attribute_group *amd64_edac_attr_groups[] = {
NULL
};

+/* Set the number of Unified Memory Controllers in the system. */
+static void compute_num_umcs(void)
+{
+ u8 model = boot_cpu_data.x86_model;
+
+ if (boot_cpu_data.x86 < 0x17)
+ return;
+
+ if (model >= 0x30 && model <= 0x3f)
+ num_umcs = 8;
+ else
+ num_umcs = 2;
+
+ edac_dbg(1, "Number of UMCs: %x", num_umcs);
+}
+
static int init_one_instance(unsigned int nid)
{
struct pci_dev *F3 = node_to_amd_nb(nid)->misc;
@@ -3276,7 +3284,7 @@ static int init_one_instance(unsigned int nid)
goto err_free;

if (pvt->fam >= 0x17) {
- pvt->umc = kcalloc(NUM_UMCS, sizeof(struct amd64_umc), GFP_KERNEL);
+ pvt->umc = kcalloc(num_umcs, sizeof(struct amd64_umc), GFP_KERNEL);
if (!pvt->umc) {
ret = -ENOMEM;
goto err_free;
@@ -3497,6 +3505,8 @@ static int __init amd64_edac_init(void)
if (!msrs)
goto err_free;

+ compute_num_umcs();
+
for (i = 0; i < amd_nb_num(); i++) {
err = probe_one_instance(i);
if (err) {
diff --git a/drivers/edac/amd64_edac.h b/drivers/edac/amd64_edac.h
index de8dbb0b42b5..40e63cea2d81 100644
--- a/drivers/edac/amd64_edac.h
+++ b/drivers/edac/amd64_edac.h
@@ -274,7 +274,11 @@

#define UMC_SDP_INIT BIT(31)

-#define NUM_UMCS 2
+/*
+ * Number of Unified Memory Controllers
+ * Set during driver init based on family/model.
+ */
+static u8 num_umcs;

enum amd_families {
K8_CPUS = 0,
@@ -399,8 +403,8 @@ struct err_info {

static inline u32 get_umc_base(u8 channel)
{
- /* ch0: 0x50000, ch1: 0x150000 */
- return 0x50000 + (!!channel << 20);
+ /* chY: 0xY50000 */
+ return 0x50000 + (channel << 20);
}

static inline u64 get_dram_base(struct amd64_pvt *pvt, u8 i)
--
2.17.1