Skip to contentSkip to conversation

some backport for stable

PATCH Applied Jack Wang 2 participants 5 messages Mar 5, 2020 – Mar 6, 2020

Subsystems (2)
lore
Share & tools
lore

Jack Wang

author
#0 lore raw

cover some backport for stable

Hi Greg, hi Sasha,

Please consider following backport for next stable release.

First patch for vhost_net has been tested with our in house regression tests with VM migration with kernel 4.14.171, but should be also applied to older stable tree.

mce patch is to fix a call trace in EPYC Rome server during boot.

EDAC one is to fix a regression backported to at least 4.14 and 4.19, should be applied to both tree, kernel 5.4/5.5 already have the patch.

Regards,
Jack Wang from IONOS SE.

Eugenio Pérez (1):
vhost: Check docket sk_family instead of call getname

Yazen Ghannam (2):
x86/mce: Handle varying MCA bank counts
EDAC/amd64: Set grain per DIMM

 arch/x86/kernel/cpu/mcheck/mce-inject.c | 14 +++++++-------
 arch/x86/kernel/cpu/mcheck/mce.c        | 22 +++++++---------------
 drivers/edac/amd64_edac.c               |  1 +
 drivers/vhost/net.c                     | 13 ++-----------
 4 files changed, 17 insertions(+), 33 deletions(-)
Signature · 1 line
2.17.1
author
#1 lore raw

PATCH 1/3 vhost: Check docket sk_family instead of call getname

From: Eugenio Pérez <[email protected]>

commit 42d84c8490f9f0931786f1623191fcab397c3d64 upstream.

Doing so, we save one call to get data we already have in the struct.

Also, since there is no guarantee that getname use sockaddr_ll parameter beyond its size, we add a little bit of security here. It should do not do beyond MAX_ADDR_LEN, but syzbot found that ax25_getname writes more (72 bytes, the size of full_sockaddr_ax25, versus 20 + 32 bytes of sockaddr_ll + MAX_ADDR_LEN in syzbot repro).

Fixes 3a4d5c94e9593 ("vhost_net: a kernel-level virtio server")
Reported-by [email protected]
[jwang: backport to 4.14]
---
 drivers/vhost/net.c | 13 ++-----------
 1 file changed, 2 insertions(+), 11 deletions(-)
drivers/vhost/net.c +2 -11
index 4d11152e60c1..8fe07622ae59 100644
@@ -1025,11 +1025,7 @@ static int vhost_net_release(struct inode *inode, struct file *f)
 
 static struct socket *get_raw_socket(int fd)
 {
-	struct {
-		struct sockaddr_ll sa;
-		char  buf[MAX_ADDR_LEN];
-	} uaddr;
-	int uaddr_len = sizeof uaddr, r;
+	int r;
 	struct socket *sock = sockfd_lookup(fd, &r);
 
 	if (!sock)
@@ -1041,12 +1037,7 @@ static struct socket *get_raw_socket(int fd)
 		goto err;
 	}
 
-	r = sock->ops->getname(sock, (struct sockaddr *)&uaddr.sa,
-			       &uaddr_len, 0);
-	if (r)
-		goto err;
-
-	if (uaddr.sa.sll_family != AF_PACKET) {
+	if (sock->sk->sk_family != AF_PACKET) {
 		r = -EPFNOSUPPORT;
 		goto err;
 	}
Signature · 1 line
2.17.1
author
#2 lore raw

PATCH 2/3 x86/mce: Handle varying MCA bank counts

In reply to Jack Wang

From: Yazen Ghannam <[email protected]>

[ Upstream commit 006c077041dc73b9490fffc4c6af5befe0687110 ]

Linux reads MCG_CAP[Count] to find the number of MCA banks visible to a CPU. Currently, this number is the same for all CPUs and a warning is shown if there is a difference. The number of banks is overwritten with the MCG_CAP[Count] value of each following CPU that boots.

According to the Intel SDM and AMD APM, the MCG_CAP[Count] value gives the number of banks that are available to a "processor implementation". The AMD BKDGs/PPRs further clarify that this value is per core. This value has historically been the same for every core in the system, but that is not an architectural requirement.

Future AMD systems may have different MCG_CAP[Count] values per core, so the assumption that all CPUs will have the same MCG_CAP[Count] value will no longer be valid.

Also, the first CPU to boot will allocate the struct mce_banks[] array using the number of banks based on its MCG_CAP[Count] value. The machine check handler and other functions use the global number of banks to iterate and index into the mce_banks[] array. So it's possible to use an out-of-bounds index on an asymmetric system where a following CPU sees a MCG_CAP[Count] value greater than its predecessors.

Thus, allocate the mce_banks[] array to the maximum number of banks. This will avoid the potential out-of-bounds index since the value of mca_cfg.banks is capped to MAX_NR_BANKS.

Set the value of mca_cfg.banks equal to the max of the previous value and the value for the current CPU. This way mca_cfg.banks will always represent the max number of banks detected on any CPU in the system.

This will ensure that all CPUs will access all the banks that are visible to them. A CPU that can access fewer than the max number of banks will find the registers of the extra banks to be read-as-zero.

Furthermore, print the resulting number of MCA banks in use. Do this in mcheck_late_init() so that the final value is printed after all CPUs have been initialized.

Finally, get bank count from target CPU when doing injection with mce-inject module.

[ bp: Remove out-of-bounds example, passify and cleanup commit message. ]

[jwang: cherry-pick to fix boot warning in arch/x86/kernel/cpu/mcheck/mce.c:1549 in epyc rome server]

---
 arch/x86/kernel/cpu/mcheck/mce-inject.c | 14 +++++++-------
 arch/x86/kernel/cpu/mcheck/mce.c        | 22 +++++++---------------
 2 files changed, 14 insertions(+), 22 deletions(-)
arch/x86/kernel/cpu/mcheck/mce-inject.c +7 -7
index f12141ba9a76..e57b59762f9f 100644
@@ -46,8 +46,6 @@
 static struct mce i_mce;
 static struct dentry *dfs_inj;
 
-static u8 n_banks;
-
 #define MAX_FLAG_OPT_SIZE	4
 #define NBCFG			0x44
 
@@ -570,9 +568,15 @@ static void do_inject(void)
 static int inj_bank_set(void *data, u64 val)
 {
 	struct mce *m = (struct mce *)data;
+	u8 n_banks;
+	u64 cap;
+
+	/* Get bank count on target CPU so we can handle non-uniform values. */
+	rdmsrl_on_cpu(m->extcpu, MSR_IA32_MCG_CAP, &cap);
+	n_banks = cap & MCG_BANKCNT_MASK;
 
 	if (val >= n_banks) {
-		pr_err("Non-existent MCE bank: %llu\n", val);
+		pr_err("MCA bank %llu non-existent on CPU%d\n", val, m->extcpu);
 		return -EINVAL;
 	}
 
@@ -665,10 +669,6 @@ static struct dfs_node {
 static int __init debugfs_init(void)
 {
 	unsigned int i;
-	u64 cap;
-
-	rdmsrl(MSR_IA32_MCG_CAP, cap);
-	n_banks = cap & MCG_BANKCNT_MASK;
 
 	dfs_inj = debugfs_create_dir("mce-inject", NULL);
 	if (!dfs_inj)
arch/x86/kernel/cpu/mcheck/mce.c +7 -15
index 0b0e44f85393..95c09db1bba2 100644
@@ -1499,13 +1499,12 @@ EXPORT_SYMBOL_GPL(mce_notify_irq);
 static int __mcheck_cpu_mce_banks_init(void)
 {
 	int i;
-	u8 num_banks = mca_cfg.banks;
 
-	mce_banks = kzalloc(num_banks * sizeof(struct mce_bank), GFP_KERNEL);
+	mce_banks = kcalloc(MAX_NR_BANKS, sizeof(struct mce_bank), GFP_KERNEL);
 	if (!mce_banks)
 		return -ENOMEM;
 
-	for (i = 0; i < num_banks; i++) {
+	for (i = 0; i < MAX_NR_BANKS; i++) {
 		struct mce_bank *b = &mce_banks[i];
 
 		b->ctl = -1ULL;
@@ -1519,28 +1518,19 @@ static int __mcheck_cpu_mce_banks_init(void)
  */
 static int __mcheck_cpu_cap_init(void)
 {
-	unsigned b;
 	u64 cap;
+	u8 b;
 
 	rdmsrl(MSR_IA32_MCG_CAP, cap);
 
 	b = cap & MCG_BANKCNT_MASK;
-	if (!mca_cfg.banks)
-		pr_info("CPU supports %d MCE banks\n", b);
-
-	if (b > MAX_NR_BANKS) {
-		pr_warn("Using only %u machine check banks out of %u\n",
-			MAX_NR_BANKS, b);
+	if (WARN_ON_ONCE(b > MAX_NR_BANKS))
 		b = MAX_NR_BANKS;
-	}
 
-	/* Don't support asymmetric configurations today */
-	WARN_ON(mca_cfg.banks != 0 && b != mca_cfg.banks);
-	mca_cfg.banks = b;
+	mca_cfg.banks = max(mca_cfg.banks, b);
 
 	if (!mce_banks) {
 		int err = __mcheck_cpu_mce_banks_init();
-
 		if (err)
 			return err;
 	}
@@ -2470,6 +2460,8 @@ EXPORT_SYMBOL_GPL(mcsafe_key);
 
 static int __init mcheck_late_init(void)
 {
+	pr_info("Using %d MCE banks\n", mca_cfg.banks);
+
 	if (mca_cfg.recovery)
 		static_branch_inc(&mcsafe_key);
 
Signature · 1 line
2.17.1
author
#3 lore raw

PATCH 3/3 EDAC/amd64: Set grain per DIMM

In reply to Jack Wang

From: Yazen Ghannam <[email protected]>

[ Upstream commit 466503d6b1b33be46ab87c6090f0ade6c6011cbc ]

The following commit introduced a warning on error reports without a non-zero grain value.

3724ace582d9 ("EDAC/mc: Fix grain_bits calculation")

The amd64_edac_mod module does not provide a value, so the warning will be given on the first reported memory error.

Set the grain per DIMM to cacheline size (64 bytes). This is the current recommendation.

Fixes 3724ace582d9 ("EDAC/mc: Fix grain_bits calculation")
Cc James Morse <[email protected]>
Cc Mauro Carvalho Chehab <[email protected]>
Cc Robert Richter <[email protected]>
Cc Tony Luck <[email protected]>
[jwang: backport to 4.14 for fix warning during memory error. ]
---
 drivers/edac/amd64_edac.c | 1 +
 1 file changed, 1 insertion(+)
drivers/edac/amd64_edac.c +1 -0
index 40fb0e7ff8fd..b36abd253786 100644
@@ -2863,6 +2863,7 @@ static int init_csrows(struct mem_ctl_info *mci)
 			dimm = csrow->channels[j]->dimm;
 			dimm->mtype = pvt->dram_type;
 			dimm->edac_mode = edac_mode;
+			dimm->grain = 64;
 		}
 	}
 
Signature · 1 line
2.17.1
maintainer Applied
#4 lore raw
In reply to Jack Wang
Quoted 23 lines · Jack Wang

Hi Greg, hi Sasha,

Please consider following backport for next stable release.

First patch for vhost_net has been tested with our in house regression tests with VM migration with kernel 4.14.171, but should be also applied to older stable tree.

mce patch is to fix a call trace in EPYC Rome server during boot.

EDAC one is to fix a regression backported to at least 4.14 and 4.19, should be applied to both tree, kernel 5.4/5.5 already have the patch.

Regards,
Jack Wang from IONOS SE.

Eugenio Pérez (1):
vhost: Check docket sk_family instead of call getname

Yazen Ghannam (2):
x86/mce: Handle varying MCA bank counts
EDAC/amd64: Set grain per DIMM

Quoted 5 lines
arch/x86/kernel/cpu/mcheck/mce-inject.c | 14 +++++++-------
arch/x86/kernel/cpu/mcheck/mce.c        | 22 +++++++---------------
drivers/edac/amd64_edac.c               |  1 +
drivers/vhost/net.c                     | 13 ++-----------
4 files changed, 17 insertions(+), 33 deletions(-)

Queued up. I took the EDAC commit to 4.19 as well.

Signature · 2 lines
Thanks,
Sasha