| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20220315095612epcms2p304482f5160ae91a120d719c88dbe368d@epcms2p3>
Date: Tue, 15 Mar 2022 18:56:12 +0900
From: Sungup Moon <sungup.moon@...sung.com>
To: "hch@....de" <hch@....de>
CC: "kbusch@...nel.org" <kbusch@...nel.org>,
"axboe@...com" <axboe@...com>,
"sagi@...mberg.me" <sagi@...mberg.me>,
"linux-nvme@...ts.infradead.org" <linux-nvme@...ts.infradead.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE:(2) [PATCH v2] driver/nvme/host: Support duplicated nsid for the
private ns
I'll answer your opinion.
1. ns_head leak issue
I don't think that is leaked ns_head. Because although all ids and nsid are same
through all namespaces, each namespaces are indenpendent namespace and each of that
should have independent data structure.
Duplicated nsid private namespace is different from the shared namespace even though
same information.
2. nvme_mpath_set_disk_name issue
Yes, I also agree that subsystem-wide IDA is very important data. However, I
implemented without nvme_mpath_set_disk_name modification at the first time, it is
hard to decide which namespace are connected to private controller.
As you know, each nvme controller start initiating at a time. So, each controller
structures are sequentially initiated, but each namespaces structures are initiated
independently because of multi-processing on cpu. So, all namespace can have different
instance number every boot-up time, and it makes hard to track and control the private
namespace on the controller or device failure.
Anyway, the private namespace is same condition with no-multipath situation (because
private namespace cannot shared between controllers) so I think that the private
namespace should follow the naming rule with no-multipath situation.
--------- Original Message ---------
Sender : hch@....de <hch@....de>
Date : 2022-03-15 16:12 (GMT+9)
Title : Re: [PATCH v2] driver/nvme/host: Support duplicated nsid for the private ns
I looked at this a bit more and found two issues:
- nvme_init_ns_head will now leak the ns_head for the private namespaces
with potentially duplicate IDs case.
- nvme_mpath_set_disk_name still needs to use the subsystem-wide IDA
for the nvme instance name as the subsystem and controller ones
could otherwise clash.
Let me know what you think of this version:
---
From 1b217962cc10fa59eae98fc112adc64bddc462b3 Mon Sep 17 00:00:00 2001
From: Sungup Moon <sungup.moon@...sung.com>
Date: Mon, 14 Mar 2022 20:05:45 +0900
Subject: nvme: allow duplicated NSIDs for the private namespaces
A NVMe subsystem with multiple controller can have private namespaces
that use the same NSID under some conditions:
"If Namespace Management, ANA Reporting, or NVM Sets are supported, the
NSIDs shall be unique within the NVM subsystem. If the Namespace
Management, ANA Reporting, and NVM Sets are not supported, then NSIDs:
a) for shared namespace shall be unique; and
b) for private namespace are not required to be unique."
Reference: Section 6.1.6 NSID and Namespace Usage; NVM Express 1.4c spec.
Make sure this specific setup is supported in Linux.
Signed-off-by: Sungup Moon <sungup.moon@...sung.com>
Signed-off-by: Christoph Hellwig <hch@....de>
---
drivers/nvme/host/core.c | 7 ++++++-
drivers/nvme/host/multipath.c | 7 ++++---
drivers/nvme/host/nvme.h | 19 +++++++++++++++++++
include/linux/nvme.h | 1 +
4 files changed, 30 insertions(+), 4 deletions(-)
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index f8084ded69e50..c7127d439b3de 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -3657,7 +3657,12 @@ static struct nvme_ns_head *nvme_find_ns_head(struct nvme_subsystem *subsys,
lockdep_assert_held(&subsys->lock);
list_for_each_entry(h, &subsys->nsheads, entry) {
- if (h->ns_id != nsid)
+ /*
+ * Private namespaces can share NSIDs under some conditions.
+ * In that case we can't use the same ns_head for namespaces
+ * with the same NSID.
+ */
+ if (h->ns_id != nsid || !nvme_is_uniqueue_nsid(ctrl, head))
continue;
if (!list_empty(&h->list) && nvme_tryget_ns_head(h))
return h;
diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
index d13b81cd6225c..6b6df1016cb91 100644
--- a/drivers/nvme/host/multipath.c
+++ b/drivers/nvme/host/multipath.c
@@ -504,10 +504,11 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head)
/*
* Add a multipath node if the subsystems supports multiple controllers.
- * We also do this for private namespaces as the namespace sharing data could
- * change after a rescan.
+ * We also do this for private namespaces as the namespace sharing flag
+ * could change after a rescan.
*/
- if (!(ctrl->subsys->cmic & NVME_CTRL_CMIC_MULTI_CTRL) || !multipath)
+ if (!(ctrl->subsys->cmic & NVME_CTRL_CMIC_MULTI_CTRL) ||
+ !nvme_is_uniqueue_nsid(ctrl, head) || !multipath)
return 0;
head->disk = blk_alloc_disk(ctrl->numa_node);
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index 587d92df118b7..9add586434929 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -718,6 +718,25 @@ static inline bool nvme_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
return queue_live;
return __nvme_check_ready(ctrl, rq, queue_live);
}
+
+/*
+ * NSID shall be unique for all shared namespaces, or if at least one of the
+ * following conditions is met:
+ * 1. Namespace Management is supported by the controller
+ * 2. ANA is supported by the controller
+ * 3. NVM Set are supported by the controller
+ *
+ * In other case, private namespace are not required to report a unique NSID.
+ */
+static inline bool nvme_is_uniqueue_nsid(struct nvme_ctrl *ctrl,
+ struct nvme_ns_head *head)
+{
+ return head->shared ||
+ (ctrl->oacs & NVME_CTRL_OACS_NS_MNGT_SUPP) ||
+ (ctrl->subsys->cmic & NVME_CTRL_CMIC_ANA) ||
+ (ctrl->ctratt & NVME_CTRL_CTRATT_NVM_SETS);
+}
+
int nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
void *buf, unsigned bufflen);
int __nvme_submit_sync_cmd(struct request_queue *q, struct nvme_command *cmd,
diff --git a/include/linux/nvme.h b/include/linux/nvme.h
index 9dbc3ef4daf7c..2dcee34d467d6 100644
--- a/include/linux/nvme.h
+++ b/include/linux/nvme.h
@@ -345,6 +345,7 @@ enum {
NVME_CTRL_ONCS_TIMESTAMP = 1 << 6,
NVME_CTRL_VWC_PRESENT = 1 << 0,
NVME_CTRL_OACS_SEC_SUPP = 1 << 0,
+ NVME_CTRL_OACS_NS_MNGT_SUPP = 1 << 3,
NVME_CTRL_OACS_DIRECTIVES = 1 << 5,
NVME_CTRL_OACS_DBBUF_SUPP = 1 << 8,
NVME_CTRL_LPA_CMD_EFFECTS_LOG = 1 << 1,
--
2.30.2
Powered by blists - more mailing lists