Skip to contentSkip to conversation

liveupdate: FLB refcounting improvements

PATCH Applied David Matlack 3 participants 12 messages Apr 23 – May 1

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover liveupdate: FLB refcounting improvements

This series can be found on GitHub:

https://github.com/dmatlack/linux/tree/liveupdate/flb/refcount/v1

This series has 2 patches to improve how FLB refcounting works in preparation for using FLB in the PCI core. The first patch converts the existing refcounting to use refcount_t, and the second patch allows users of the incoming FLB to hold references to the FLB to avoid it getting freed.

Users of the outgoing FLB don't need to take an extra reference yet since it's always done in the context of a preserve/unpreserve file callback, which are already synchronized with outgoing FLB creation and destruction.

This series has been tested against the existing Live Update selftests (liveupdate, luo_kexec_simple, and luo_multi_session) on an Intel EMR host.

David Matlack (2):
liveupdate: Use refcount_t for FLB reference counts
liveupdate: Reference count incoming FLB data

 include/linux/liveupdate.h  |  9 ++++++-
 kernel/liveupdate/luo_flb.c | 52 ++++++++++++++++++-------------------
 lib/tests/liveupdate.c      |  3 +++
 3 files changed, 37 insertions(+), 27 deletions(-)

base-commit: 2a4c0c11c0193889446cdb6f1540cc2b9aff97dd

Signature · 1 line
2.54.0.rc2.544.gc7ae2d5bb8-goog
#1 lore raw

PATCH 1/2 liveupdate: Use refcount_t for FLB reference counts

Use refcount_t instead of a raw integer to keep track of references on incoming and outgoing FLBs. Using refcount_t provides protection from overflow, underflow, and other issues.

Fixes cab056f2aae7 ("liveupdate: luo_flb: introduce File-Lifecycle-Bound global state")
---
 include/linux/liveupdate.h  |  3 ++-
 kernel/liveupdate/luo_flb.c | 22 ++++++++++------------
 2 files changed, 12 insertions(+), 13 deletions(-)
include/linux/liveupdate.h +2 -1
index 30c5a39ff9e9..8d3bbc35c828 100644
@@ -12,6 +12,7 @@
 #include <linux/kho/abi/luo.h>
 #include <linux/list.h>
 #include <linux/mutex.h>
+#include <linux/refcount.h>
 #include <linux/rwsem.h>
 #include <linux/types.h>
 #include <uapi/linux/liveupdate.h>
@@ -175,7 +176,7 @@ struct liveupdate_flb_ops {
  * @retrieved: True once the FLB's retrieve() callback has run.
  */
 struct luo_flb_private_state {
-	long count;
+	refcount_t count;
 	u64 data;
 	void *obj;
 	struct mutex lock;
kernel/liveupdate/luo_flb.c +10 -12
index 00f5494812c4..59c5f31ab767 100644
@@ -111,7 +111,7 @@ static int luo_flb_file_preserve_one(struct liveupdate_flb *flb)
 	struct luo_flb_private *private = luo_flb_get_private(flb);
 
 	scoped_guard(mutex, &private->outgoing.lock) {
-		if (!private->outgoing.count) {
+		if (!refcount_read(&private->outgoing.count)) {
 			struct liveupdate_flb_op_args args = {0};
 			int err;
 
@@ -126,8 +126,10 @@ static int luo_flb_file_preserve_one(struct liveupdate_flb *flb)
 			}
 			private->outgoing.data = args.data;
 			private->outgoing.obj = args.obj;
+			refcount_set(&private->outgoing.count, 1);
+		} else {
+			refcount_inc(&private->outgoing.count);
 		}
-		private->outgoing.count++;
 	}
 
 	return 0;
@@ -138,8 +140,7 @@ static void luo_flb_file_unpreserve_one(struct liveupdate_flb *flb)
 	struct luo_flb_private *private = luo_flb_get_private(flb);
 
 	scoped_guard(mutex, &private->outgoing.lock) {
-		private->outgoing.count--;
-		if (!private->outgoing.count) {
+		if (refcount_dec_and_test(&private->outgoing.count)) {
 			struct liveupdate_flb_op_args args = {0};
 
 			args.flb = flb;
@@ -178,7 +179,7 @@ static int luo_flb_retrieve_one(struct liveupdate_flb *flb)
 	for (int i = 0; i < fh->header_ser->count; i++) {
 		if (!strcmp(fh->ser[i].name, flb->compatible)) {
 			private->incoming.data = fh->ser[i].data;
-			private->incoming.count = fh->ser[i].count;
+			refcount_set(&private->incoming.count, fh->ser[i].count);
 			found = true;
 			break;
 		}
@@ -208,12 +209,8 @@ static int luo_flb_retrieve_one(struct liveupdate_flb *flb)
 static void luo_flb_file_finish_one(struct liveupdate_flb *flb)
 {
 	struct luo_flb_private *private = luo_flb_get_private(flb);
-	u64 count;
 
-	scoped_guard(mutex, &private->incoming.lock)
-		count = --private->incoming.count;
-
-	if (!count) {
+	if (refcount_dec_and_test(&private->incoming.count)) {
 		struct liveupdate_flb_op_args args = {0};
 
 		if (!private->incoming.retrieved) {
@@ -652,12 +649,13 @@ void luo_flb_serialize(void)
 	guard(rwsem_read)(&luo_register_rwlock);
 	list_private_for_each_entry(gflb, &luo_flb_global.list, private.list) {
 		struct luo_flb_private *private = luo_flb_get_private(gflb);
+		long count = refcount_read(&private->outgoing.count);
 
-		if (private->outgoing.count > 0) {
+		if (count > 0) {
 			strscpy(fh->ser[i].name, gflb->compatible,
 				sizeof(fh->ser[i].name));
 			fh->ser[i].data = private->outgoing.data;
-			fh->ser[i].count = private->outgoing.count;
+			fh->ser[i].count = count;
 			i++;
 		}
 	}
Signature · 1 line
2.54.0.rc2.544.gc7ae2d5bb8-goog
maintainer
#2 lore raw

Re: [PATCH 1/2] liveupdate: Use refcount_t for FLB reference counts

Quoted 46 lines · David Matlack

Use refcount_t instead of a raw integer to keep track of references on incoming and outgoing FLBs. Using refcount_t provides protection from overflow, underflow, and other issues.

Fixes cab056f2aae7 ("liveupdate: luo_flb: introduce File-Lifecycle-Bound global state")
---
 include/linux/liveupdate.h  |  3 ++-
 kernel/liveupdate/luo_flb.c | 22 ++++++++++------------
 2 files changed, 12 insertions(+), 13 deletions(-)
include/linux/liveupdate.h +2 -1
index 30c5a39ff9e9..8d3bbc35c828 100644
@@ -12,6 +12,7 @@
 #include <linux/kho/abi/luo.h>
 #include <linux/list.h>
 #include <linux/mutex.h>
+#include <linux/refcount.h>
 #include <linux/rwsem.h>
 #include <linux/types.h>
 #include <uapi/linux/liveupdate.h>
@@ -175,7 +176,7 @@ struct liveupdate_flb_ops {
  * @retrieved: True once the FLB's retrieve() callback has run.
  */
 struct luo_flb_private_state {
-	long count;
+	refcount_t count;
 	u64 data;
 	void *obj;
 	struct mutex lock;
kernel/liveupdate/luo_flb.c +1 -1
index 00f5494812c4..59c5f31ab767 100644
@@ -111,7 +111,7 @@ static int luo_flb_file_preserve_one(struct liveupdate_flb *flb)
 	struct luo_flb_private *private = luo_flb_get_private(flb);
 
 	scoped_guard(mutex, &private->outgoing.lock) {
-		if (!private->outgoing.count) {
+		if (!refcount_read(&private->outgoing.count)) {
 			struct liveupdate_flb_op_args args = {0};
 			int err;
 

kernel/liveupdate/luo_flb.c static int luo_flb_file_preserve_one(struct liveupdate_flb *flb) · line 130

Quoted 8 lines
kernel/liveupdate/luo_flb.c +3 -1
@@ -126,8 +126,10 @@ static int luo_flb_file_preserve_one(struct liveupdate_flb *flb)
 			}
 			private->outgoing.data = args.data;
 			private->outgoing.obj = args.obj;
+			refcount_set(&private->outgoing.count, 1);
+		} else {
+			refcount_inc(&private->outgoing.count);
 		}
-		private->outgoing.count++;

It should be: refcount_inc(&private->outgoing.count); for both cases, as it was before.

Additionally, please add refcount_set(&private->outgoing.count, 0) to luo_flb_get_private, where the rest of the private fields are initialized.

In general, I prefer to avoid refcount_set() because it breaks continuity and makes debugging hard. It should only be used only during initialization, so moving it to where the other fields are initialized is the correct approach.

Quoted 55 lines
kernel/liveupdate/luo_flb.c +6 -10
 	}
 
 	return 0;
@@ -138,8 +140,7 @@ static void luo_flb_file_unpreserve_one(struct liveupdate_flb *flb)
 	struct luo_flb_private *private = luo_flb_get_private(flb);
 
 	scoped_guard(mutex, &private->outgoing.lock) {
-		private->outgoing.count--;
-		if (!private->outgoing.count) {
+		if (refcount_dec_and_test(&private->outgoing.count)) {
 			struct liveupdate_flb_op_args args = {0};
 
 			args.flb = flb;
@@ -178,7 +179,7 @@ static int luo_flb_retrieve_one(struct liveupdate_flb *flb)
 	for (int i = 0; i < fh->header_ser->count; i++) {
 		if (!strcmp(fh->ser[i].name, flb->compatible)) {
 			private->incoming.data = fh->ser[i].data;
-			private->incoming.count = fh->ser[i].count;
+			refcount_set(&private->incoming.count, fh->ser[i].count);
 			found = true;
 			break;
 		}
@@ -208,12 +209,8 @@ static int luo_flb_retrieve_one(struct liveupdate_flb *flb)
 static void luo_flb_file_finish_one(struct liveupdate_flb *flb)
 {
 	struct luo_flb_private *private = luo_flb_get_private(flb);
-	u64 count;
 
-	scoped_guard(mutex, &private->incoming.lock)
-		count = --private->incoming.count;
-
-	if (!count) {
+	if (refcount_dec_and_test(&private->incoming.count)) {
 		struct liveupdate_flb_op_args args = {0};
 
 		if (!private->incoming.retrieved) {
@@ -652,12 +649,13 @@ void luo_flb_serialize(void)
 	guard(rwsem_read)(&luo_register_rwlock);
 	list_private_for_each_entry(gflb, &luo_flb_global.list, private.list) {
 		struct luo_flb_private *private = luo_flb_get_private(gflb);
+		long count = refcount_read(&private->outgoing.count);
 
-		if (private->outgoing.count > 0) {
+		if (count > 0) {
 			strscpy(fh->ser[i].name, gflb->compatible,
 				sizeof(fh->ser[i].name));
 			fh->ser[i].data = private->outgoing.data;
-			fh->ser[i].count = private->outgoing.count;
+			fh->ser[i].count = count;
 			i++;
 		}
 	}
Signature · 1 line
2.54.0.rc2.544.gc7ae2d5bb8-goog
#3 lore raw

Re: [PATCH 1/2] liveupdate: Use refcount_t for FLB reference counts

Quoted 14 lines · Pasha Tatashin

kernel/liveupdate/luo_flb.c static int luo_flb_file_preserve_one(struct liveupdate_flb *flb) · line 130

Quoted 9 lines · David Matlack
kernel/liveupdate/luo_flb.c +3 -1
@@ -126,8 +126,10 @@ static int luo_flb_file_preserve_one(struct liveupdate_flb *flb)
                      }
                      private->outgoing.data = args.data;
                      private->outgoing.obj = args.obj;
+                     refcount_set(&private->outgoing.count, 1);
+             } else {
+                     refcount_inc(&private->outgoing.count);
              }
-             private->outgoing.count++;

It should be: refcount_inc(&private->outgoing.count); for both cases, as it was before.

Calling refcount_inc() when the refcount is 0 triggers the REFCOUNT_ADD_UAF warning.

Quoted 3 lines

Additionally, please add refcount_set(&private->outgoing.count, 0) to luo_flb_get_private, where the rest of the private fields are initialized.

Will do.

Quoted 4 lines

In general, I prefer to avoid refcount_set() because it breaks continuity and makes debugging hard. It should only be used only during initialization, so moving it to where the other fields are initialized is the correct approach.

maintainer Reviewed-by
#4 lore raw

Re: [PATCH 1/2] liveupdate: Use refcount_t for FLB reference counts

Quoted 17 lines · David Matlack
Quoted 14 lines · Pasha Tatashin

kernel/liveupdate/luo_flb.c static int luo_flb_file_preserve_one(struct liveupdate_flb *flb) · line 130

Quoted 9 lines · David Matlack
kernel/liveupdate/luo_flb.c +3 -1
@@ -126,8 +126,10 @@ static int luo_flb_file_preserve_one(struct liveupdate_flb *flb)
                      }
                      private->outgoing.data = args.data;
                      private->outgoing.obj = args.obj;
+                     refcount_set(&private->outgoing.count, 1);
+             } else {
+                     refcount_inc(&private->outgoing.count);
              }
-             private->outgoing.count++;

It should be: refcount_inc(&private->outgoing.count); for both cases, as it was before.

Quoted 2 lines

Calling refcount_inc() when the refcount is 0 triggers the REFCOUNT_ADD_UAF warning.

Oh you are right, sigh, in this case what you have is OK.

Quoted 6 lines
Quoted 4 lines

Additionally, please add refcount_set(&private->outgoing.count, 0) to luo_flb_get_private, where the rest of the private fields are initialized.

Will do.

No need to do this...

Quoted 5 lines
Quoted 5 lines

In general, I prefer to avoid refcount_set() because it breaks continuity and makes debugging hard. It should only be used only during initialization, so moving it to where the other fields are initialized is the correct approach.

Reviewed-by on v1 1/2
#5 lore raw

Re: [PATCH 1/2] liveupdate: Use refcount_t for FLB reference counts

In reply to David Matlack
Quoted 11 lines · David Matlack

Use refcount_t instead of a raw integer to keep track of references on incoming and outgoing FLBs. Using refcount_t provides protection from overflow, underflow, and other issues.

Fixes cab056f2aae7 ("liveupdate: luo_flb: introduce File-Lifecycle-Bound global state")

---
include/linux/liveupdate.h | 3 ++- kernel/liveupdate/luo_flb.c | 22 ++++++++++------------

2 files changed, 12 insertions(+), 13 deletions(-)
#6 lore raw

PATCH 2/2 liveupdate: Reference count incoming FLB data

In reply to David Matlack

Increment the incoming FLB refcount in liveupdate_flb_get_incoming() so that the FLB structure cannot be freed while the caller is actively using it. Add an additional liveupdate_flb_put_incoming() function so the caller can explicitly indicate when it is done using the FLB data.

During a Live Update, a subsystem might need to hold onto the incoming File-Lifecycle-Bound (FLB) data for an extended period, such as during device enumeration. Incrementing the reference count guarantees that the data remains valid and accessible until the subsystem releases it, preventing future use-after-free bugs.

Fixes cab056f2aae7 ("liveupdate: luo_flb: introduce File-Lifecycle-Bound global state")
---
 include/linux/liveupdate.h  |  6 ++++++
 kernel/liveupdate/luo_flb.c | 32 +++++++++++++++++---------------
 lib/tests/liveupdate.c      |  3 +++
 3 files changed, 26 insertions(+), 15 deletions(-)
include/linux/liveupdate.h +6 -0
index 8d3bbc35c828..88722e5caf02 100644
@@ -240,6 +240,8 @@ void liveupdate_unregister_flb(struct liveupdate_file_handler *fh,
 			       struct liveupdate_flb *flb);
 
 int liveupdate_flb_get_incoming(struct liveupdate_flb *flb, void **objp);
+void liveupdate_flb_put_incoming(struct liveupdate_flb *flb);
+
 int liveupdate_flb_get_outgoing(struct liveupdate_flb *flb, void **objp);
 
 #else /* CONFIG_LIVEUPDATE */
@@ -280,6 +282,10 @@ static inline int liveupdate_flb_get_incoming(struct liveupdate_flb *flb,
 	return -EOPNOTSUPP;
 }
 
+static inline void liveupdate_flb_put_incoming(struct liveupdate_flb *flb)
+{
+}
+
 static inline int liveupdate_flb_get_outgoing(struct liveupdate_flb *flb,
 					      void **objp)
 {
kernel/liveupdate/luo_flb.c +17 -15
index 59c5f31ab767..8f5c5dd01cd0 100644
@@ -165,7 +165,7 @@ static int luo_flb_retrieve_one(struct liveupdate_flb *flb)
 	bool found = false;
 	int err;
 
-	guard(mutex)(&private->incoming.lock);
+	lockdep_assert_held(&private->incoming.lock);
 
 	if (private->incoming.finished)
 		return -ENODATA;
@@ -206,12 +206,14 @@ static int luo_flb_retrieve_one(struct liveupdate_flb *flb)
 	return 0;
 }
 
-static void luo_flb_file_finish_one(struct liveupdate_flb *flb)
+void liveupdate_flb_put_incoming(struct liveupdate_flb *flb)
 {
 	struct luo_flb_private *private = luo_flb_get_private(flb);
+	struct liveupdate_flb_op_args args = {0};
 
-	if (refcount_dec_and_test(&private->incoming.count)) {
-		struct liveupdate_flb_op_args args = {0};
+	scoped_guard(mutex, &private->incoming.lock) {
+		if (!refcount_dec_and_test(&private->incoming.count))
+			return;
 
 		if (!private->incoming.retrieved) {
 			int err = luo_flb_retrieve_one(flb);
@@ -220,16 +222,14 @@ static void luo_flb_file_finish_one(struct liveupdate_flb *flb)
 				return;
 		}
 
-		scoped_guard(mutex, &private->incoming.lock) {
-			args.flb = flb;
-			args.obj = private->incoming.obj;
-			flb->ops->finish(&args);
+		args.flb = flb;
+		args.obj = private->incoming.obj;
+		flb->ops->finish(&args);
 
-			private->incoming.data = 0;
-			private->incoming.obj = NULL;
-			private->incoming.finished = true;
-			module_put(flb->ops->owner);
-		}
+		private->incoming.data = 0;
+		private->incoming.obj = NULL;
+		private->incoming.finished = true;
+		module_put(flb->ops->owner);
 	}
 }
 
@@ -312,7 +312,7 @@ void luo_flb_file_finish(struct liveupdate_file_handler *fh)
 
 	guard(rwsem_read)(&luo_register_rwlock);
 	list_for_each_entry_reverse(iter, flb_list, list)
-		luo_flb_file_finish_one(iter->flb);
+		liveupdate_flb_put_incoming(iter->flb);
 }
 
 static void luo_flb_unregister_one(struct liveupdate_file_handler *fh,
@@ -509,6 +509,8 @@ int liveupdate_flb_get_incoming(struct liveupdate_flb *flb, void **objp)
 	if (!liveupdate_enabled())
 		return -EOPNOTSUPP;
 
+	guard(mutex)(&private->incoming.lock);
+
 	if (!private->incoming.obj) {
 		int err = luo_flb_retrieve_one(flb);
 
@@ -516,7 +518,7 @@ int liveupdate_flb_get_incoming(struct liveupdate_flb *flb, void **objp)
 			return err;
 	}
 
-	guard(mutex)(&private->incoming.lock);
+	refcount_inc(&private->incoming.count);
 	*objp = private->incoming.obj;
 
 	return 0;
lib/tests/liveupdate.c +3 -0
index e4b0ecbee32f..4c08a7c6fb78 100644
@@ -105,6 +105,9 @@ static void liveupdate_test_init(void)
 			pr_err("liveupdate_flb_get_incoming for %s failed: %pe\n",
 			       flb->compatible, ERR_PTR(err));
 		}
+
+		if (!err)
+			liveupdate_flb_put_incoming(flb);
 	}
 	initialized = true;
 }
Signature · 1 line
2.54.0.rc2.544.gc7ae2d5bb8-goog
maintainer Reviewed-by on v1 2/2
#7 lore raw

Re: [PATCH 2/2] liveupdate: Reference count incoming FLB data

Quoted 5 lines · David Matlack

Increment the incoming FLB refcount in liveupdate_flb_get_incoming() so that the FLB structure cannot be freed while the caller is actively using it. Add an additional liveupdate_flb_put_incoming() function so the caller can explicitly indicate when it is done using the FLB data.

Quoted 8 lines

During a Live Update, a subsystem might need to hold onto the incoming File-Lifecycle-Bound (FLB) data for an extended period, such as during device enumeration. Incrementing the reference count guarantees that the data remains valid and accessible until the subsystem releases it, preventing future use-after-free bugs.

Fixes cab056f2aae7 ("liveupdate: luo_flb: introduce File-Lifecycle-Bound global state")

Thanks!

Pasha

Reviewed-by on v1 2/2
#8 lore raw

Re: [PATCH 2/2] liveupdate: Reference count incoming FLB data

In reply to David Matlack
Quoted 142 lines · David Matlack

Increment the incoming FLB refcount in liveupdate_flb_get_incoming() so that the FLB structure cannot be freed while the caller is actively using it. Add an additional liveupdate_flb_put_incoming() function so the caller can explicitly indicate when it is done using the FLB data.

During a Live Update, a subsystem might need to hold onto the incoming File-Lifecycle-Bound (FLB) data for an extended period, such as during device enumeration. Incrementing the reference count guarantees that the data remains valid and accessible until the subsystem releases it, preventing future use-after-free bugs.

Fixes cab056f2aae7 ("liveupdate: luo_flb: introduce File-Lifecycle-Bound global state")

---

include/linux/liveupdate.h  |  6 ++++++
kernel/liveupdate/luo_flb.c | 32 +++++++++++++++++---------------
lib/tests/liveupdate.c      |  3 +++
3 files changed, 26 insertions(+), 15 deletions(-)
include/linux/liveupdate.h +6 -0
index 8d3bbc35c828..88722e5caf02 100644
@@ -240,6 +240,8 @@ void liveupdate_unregister_flb(struct liveupdate_file_handler *fh,
 			       struct liveupdate_flb *flb);
 
 int liveupdate_flb_get_incoming(struct liveupdate_flb *flb, void **objp);
+void liveupdate_flb_put_incoming(struct liveupdate_flb *flb);
+
 int liveupdate_flb_get_outgoing(struct liveupdate_flb *flb, void **objp);
 
 #else /* CONFIG_LIVEUPDATE */
@@ -280,6 +282,10 @@ static inline int liveupdate_flb_get_incoming(struct liveupdate_flb *flb,
 	return -EOPNOTSUPP;
 }
 
+static inline void liveupdate_flb_put_incoming(struct liveupdate_flb *flb)
+{
+}
+
 static inline int liveupdate_flb_get_outgoing(struct liveupdate_flb *flb,
 					      void **objp)
 {
kernel/liveupdate/luo_flb.c +17 -15
index 59c5f31ab767..8f5c5dd01cd0 100644
@@ -165,7 +165,7 @@ static int luo_flb_retrieve_one(struct liveupdate_flb *flb)
 	bool found = false;
 	int err;
 
-	guard(mutex)(&private->incoming.lock);
+	lockdep_assert_held(&private->incoming.lock);
 
 	if (private->incoming.finished)
 		return -ENODATA;
@@ -206,12 +206,14 @@ static int luo_flb_retrieve_one(struct liveupdate_flb *flb)
 	return 0;
 }
 
-static void luo_flb_file_finish_one(struct liveupdate_flb *flb)
+void liveupdate_flb_put_incoming(struct liveupdate_flb *flb)
 {
 	struct luo_flb_private *private = luo_flb_get_private(flb);
+	struct liveupdate_flb_op_args args = {0};
 
-	if (refcount_dec_and_test(&private->incoming.count)) {
-		struct liveupdate_flb_op_args args = {0};
+	scoped_guard(mutex, &private->incoming.lock) {
+		if (!refcount_dec_and_test(&private->incoming.count))
+			return;
 
 		if (!private->incoming.retrieved) {
 			int err = luo_flb_retrieve_one(flb);
@@ -220,16 +222,14 @@ static void luo_flb_file_finish_one(struct liveupdate_flb *flb)
 				return;
 		}
 
-		scoped_guard(mutex, &private->incoming.lock) {
-			args.flb = flb;
-			args.obj = private->incoming.obj;
-			flb->ops->finish(&args);
+		args.flb = flb;
+		args.obj = private->incoming.obj;
+		flb->ops->finish(&args);
 
-			private->incoming.data = 0;
-			private->incoming.obj = NULL;
-			private->incoming.finished = true;
-			module_put(flb->ops->owner);
-		}
+		private->incoming.data = 0;
+		private->incoming.obj = NULL;
+		private->incoming.finished = true;
+		module_put(flb->ops->owner);
 	}
 }
 
@@ -312,7 +312,7 @@ void luo_flb_file_finish(struct liveupdate_file_handler *fh)
 
 	guard(rwsem_read)(&luo_register_rwlock);
 	list_for_each_entry_reverse(iter, flb_list, list)
-		luo_flb_file_finish_one(iter->flb);
+		liveupdate_flb_put_incoming(iter->flb);
 }
 
 static void luo_flb_unregister_one(struct liveupdate_file_handler *fh,
@@ -509,6 +509,8 @@ int liveupdate_flb_get_incoming(struct liveupdate_flb *flb, void **objp)
 	if (!liveupdate_enabled())
 		return -EOPNOTSUPP;
 
+	guard(mutex)(&private->incoming.lock);
+
 	if (!private->incoming.obj) {
 		int err = luo_flb_retrieve_one(flb);
 
@@ -516,7 +518,7 @@ int liveupdate_flb_get_incoming(struct liveupdate_flb *flb, void **objp)
 			return err;
 	}
 
-	guard(mutex)(&private->incoming.lock);
+	refcount_inc(&private->incoming.count);
 	*objp = private->incoming.obj;
 
 	return 0;
lib/tests/liveupdate.c +3 -0
index e4b0ecbee32f..4c08a7c6fb78 100644
@@ -105,6 +105,9 @@ static void liveupdate_test_init(void)
 			pr_err("liveupdate_flb_get_incoming for %s failed: %pe\n",
 			       flb->compatible, ERR_PTR(err));
 		}
+
+		if (!err)
+			liveupdate_flb_put_incoming(flb);
 	}
 	initialized = true;
 }
Signature · 1 line
2.54.0.rc2.544.gc7ae2d5bb8-goog
maintainer
#9 lore raw
In reply to David Matlack

Please include [email protected] so the patches are tested by Sashiko, and added to patchwork that we are using to maintaine live update work.

Pasha

Quoted 34 lines · David Matlack

This series can be found on GitHub:

https://github.com/dmatlack/linux/tree/liveupdate/flb/refcount/v1

This series has 2 patches to improve how FLB refcounting works in preparation for using FLB in the PCI core. The first patch converts the existing refcounting to use refcount_t, and the second patch allows users of the incoming FLB to hold references to the FLB to avoid it getting freed.

Users of the outgoing FLB don't need to take an extra reference yet since it's always done in the context of a preserve/unpreserve file callback, which are already synchronized with outgoing FLB creation and destruction.

This series has been tested against the existing Live Update selftests (liveupdate, luo_kexec_simple, and luo_multi_session) on an Intel EMR host.

David Matlack (2):
liveupdate: Use refcount_t for FLB reference counts
liveupdate: Reference count incoming FLB data

 include/linux/liveupdate.h  |  9 ++++++-
 kernel/liveupdate/luo_flb.c | 52 ++++++++++++++++++-------------------
 lib/tests/liveupdate.c      |  3 +++
 3 files changed, 37 insertions(+), 27 deletions(-)

base-commit: 2a4c0c11c0193889446cdb6f1540cc2b9aff97dd

Signature · 1 line
2.54.0.rc2.544.gc7ae2d5bb8-goog
#10 lore raw
Quoted 4 lines · Pasha Tatashin

Please include [email protected] so the patches are tested by Sashiko, and added to patchwork that we are using to maintaine live update work.

Sure. Should we add [email protected] to "LIVE UPDATE" in MAINTAINERS then? That way scripts/get_maintainers.pl works.

maintainer Applied
#11 lore raw
In reply to David Matlack
Quoted 11 lines · David Matlack

This series can be found on GitHub:

https://github.com/dmatlack/linux/tree/liveupdate/flb/refcount/v1

This series has 2 patches to improve how FLB refcounting works in preparation for using FLB in the PCI core. The first patch converts the existing refcounting to use refcount_t, and the second patch allows users of the incoming FLB to hold references to the FLB to avoid it getting freed.

[...]

Applied, thanks!

[1/2] liveupdate: Use refcount_t for FLB reference counts

      commit: 710153dabf699840887bc4f58e8a3da5afdbc607
[2/2] liveupdate: Reference count incoming FLB data
      commit: 74d1099bb4edfca3ce6edc61dbacbed96720034c

Best regards,

Signature · 1 line
Pasha Tatashin <[email protected]>