Skip to contentSkip to conversation

PCI/doe: Fix work struct declaration

PATCH In review [email protected] 5 participants 11 messages Nov 15, 2022 – Nov 16, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 PCI/doe: Fix work struct declaration

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

---
 drivers/cxl/core/pci.c  |  2 ++
 drivers/pci/doe.c       | 14 ++++++++++++--
 include/linux/pci-doe.h |  8 +++++---
 3 files changed, 19 insertions(+), 5 deletions(-)
drivers/cxl/core/pci.c +2 -0
index 9240df53ed87..a19c1fa0e2f4 100644
@@ -525,6 +525,7 @@ static int cxl_cdat_get_length(struct device *dev,
 	DECLARE_CDAT_DOE_TASK(CDAT_DOE_REQ(0), t);
 	int rc;
 
+	pci_doe_init_task(cdat_doe, &t.task, true);
 	rc = pci_doe_submit_task(cdat_doe, &t.task);
 	if (rc < 0) {
 		dev_err(dev, "DOE submit failed: %d", rc);
@@ -554,6 +555,7 @@ static int cxl_cdat_read_table(struct device *dev,
 		u32 *entry;
 		int rc;
 
+		pci_doe_init_task(cdat_doe, &t.task, true);
 		rc = pci_doe_submit_task(cdat_doe, &t.task);
 		if (rc < 0) {
 			dev_err(dev, "DOE submit failed: %d", rc);
drivers/pci/doe.c +12 -2
index e402f05068a5..cabeae4ae955 100644
@@ -319,6 +319,7 @@ static int pci_doe_discovery(struct pci_doe_mb *doe_mb, u8 *index, u16 *vid,
 	};
 	int rc;
 
+	pci_doe_init_task(doe_mb, &task, true);
 	rc = pci_doe_submit_task(doe_mb, &task);
 	if (rc < 0)
 		return rc;
@@ -495,6 +496,14 @@ bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type)
 }
 EXPORT_SYMBOL_GPL(pci_doe_supports_prot);
 
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack)
+{
+	task->doe_mb = doe_mb;
+	__INIT_WORK(&task->work, doe_statemachine_work, onstack);
+}
+EXPORT_SYMBOL_GPL(pci_doe_init_task);
+
 /**
  * pci_doe_submit_task() - Submit a task to be processed by the state machine
  *
@@ -517,6 +526,9 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (!pci_doe_supports_prot(doe_mb, task->prot.vid, task->prot.type))
 		return -EINVAL;
 
+	if (WARN_ON_ONCE(task->work.func != doe_statemachine_work))
+		return -EINVAL;
+
 	/*
 	 * DOE requests must be a whole number of DW and the response needs to
 	 * be big enough for at least 1 DW
@@ -528,8 +540,6 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (test_bit(PCI_DOE_FLAG_DEAD, &doe_mb->flags))
 		return -EIO;
 
-	task->doe_mb = doe_mb;
-	INIT_WORK(&task->work, doe_statemachine_work);
 	queue_work(doe_mb->work_queue, &task->work);
 	return 0;
 }
include/linux/pci-doe.h +5 -3
index ed9b4df792b8..457fc0e53d64 100644
@@ -31,8 +31,8 @@ struct pci_doe_mb;
  * @rv: Return value.  Length of received response or error (bytes)
  * @complete: Called when task is complete
  * @private: Private data for the consumer
- * @work: Used internally by the mailbox
- * @doe_mb: Used internally by the mailbox
+ * @work: Used internally by the mailbox [see pci_doe_init_task()]
+ * @doe_mb: Used internally by the mailbox [see pci_doe_init_task()]
  *
  * The payload sizes and rv are specified in bytes with the following
  * restrictions concerning the protocol.
@@ -53,7 +53,7 @@ struct pci_doe_task {
 	void (*complete)(struct pci_doe_task *task);
 	void *private;
 
-	/* No need for the user to initialize these fields */
+	/* Call pci_doe_init_task() for these */
 	struct work_struct work;
 	struct pci_doe_mb *doe_mb;
 };
@@ -72,6 +72,8 @@ struct pci_doe_task {
 
 struct pci_doe_mb *pcim_doe_create_mb(struct pci_dev *pdev, u16 cap_offset);
 bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type);
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack);
 int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task);
 
 #endif

base-commit: 30a0b95b1335e12efef89dd78518ed3e4a71a763

Signature · 1 line
2.37.2
Reviewed-by on v1 1/1
#1 lore raw
Quoted 20 lines · On Mon, 14 Nov 2022 17:19:43 -0800 [email protected] wrote:

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

Quoted 8 lines

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

Looks like a good solution to me.

Quoted 107 lines
---
 drivers/cxl/core/pci.c  |  2 ++
 drivers/pci/doe.c       | 14 ++++++++++++--
 include/linux/pci-doe.h |  8 +++++---
 3 files changed, 19 insertions(+), 5 deletions(-)
drivers/cxl/core/pci.c +2 -0
index 9240df53ed87..a19c1fa0e2f4 100644
@@ -525,6 +525,7 @@ static int cxl_cdat_get_length(struct device *dev,
 	DECLARE_CDAT_DOE_TASK(CDAT_DOE_REQ(0), t);
 	int rc;
 
+	pci_doe_init_task(cdat_doe, &t.task, true);
 	rc = pci_doe_submit_task(cdat_doe, &t.task);
 	if (rc < 0) {
 		dev_err(dev, "DOE submit failed: %d", rc);
@@ -554,6 +555,7 @@ static int cxl_cdat_read_table(struct device *dev,
 		u32 *entry;
 		int rc;
 
+		pci_doe_init_task(cdat_doe, &t.task, true);
 		rc = pci_doe_submit_task(cdat_doe, &t.task);
 		if (rc < 0) {
 			dev_err(dev, "DOE submit failed: %d", rc);
drivers/pci/doe.c +12 -2
index e402f05068a5..cabeae4ae955 100644
@@ -319,6 +319,7 @@ static int pci_doe_discovery(struct pci_doe_mb *doe_mb, u8 *index, u16 *vid,
 	};
 	int rc;
 
+	pci_doe_init_task(doe_mb, &task, true);
 	rc = pci_doe_submit_task(doe_mb, &task);
 	if (rc < 0)
 		return rc;
@@ -495,6 +496,14 @@ bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type)
 }
 EXPORT_SYMBOL_GPL(pci_doe_supports_prot);
 
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack)
+{
+	task->doe_mb = doe_mb;
+	__INIT_WORK(&task->work, doe_statemachine_work, onstack);
+}
+EXPORT_SYMBOL_GPL(pci_doe_init_task);
+
 /**
  * pci_doe_submit_task() - Submit a task to be processed by the state machine
  *
@@ -517,6 +526,9 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (!pci_doe_supports_prot(doe_mb, task->prot.vid, task->prot.type))
 		return -EINVAL;
 
+	if (WARN_ON_ONCE(task->work.func != doe_statemachine_work))
+		return -EINVAL;
+
 	/*
 	 * DOE requests must be a whole number of DW and the response needs to
 	 * be big enough for at least 1 DW
@@ -528,8 +540,6 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (test_bit(PCI_DOE_FLAG_DEAD, &doe_mb->flags))
 		return -EIO;
 
-	task->doe_mb = doe_mb;
-	INIT_WORK(&task->work, doe_statemachine_work);
 	queue_work(doe_mb->work_queue, &task->work);
 	return 0;
 }
include/linux/pci-doe.h +5 -3
index ed9b4df792b8..457fc0e53d64 100644
@@ -31,8 +31,8 @@ struct pci_doe_mb;
  * @rv: Return value.  Length of received response or error (bytes)
  * @complete: Called when task is complete
  * @private: Private data for the consumer
- * @work: Used internally by the mailbox
- * @doe_mb: Used internally by the mailbox
+ * @work: Used internally by the mailbox [see pci_doe_init_task()]
+ * @doe_mb: Used internally by the mailbox [see pci_doe_init_task()]
  *
  * The payload sizes and rv are specified in bytes with the following
  * restrictions concerning the protocol.
@@ -53,7 +53,7 @@ struct pci_doe_task {
 	void (*complete)(struct pci_doe_task *task);
 	void *private;
 
-	/* No need for the user to initialize these fields */
+	/* Call pci_doe_init_task() for these */
 	struct work_struct work;
 	struct pci_doe_mb *doe_mb;
 };
@@ -72,6 +72,8 @@ struct pci_doe_task {
 
 struct pci_doe_mb *pcim_doe_create_mb(struct pci_dev *pdev, u16 cap_offset);
 bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type);
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack);
 int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task);
 
 #endif

base-commit: 30a0b95b1335e12efef89dd78518ed3e4a71a763

#2 lore raw
In reply to [email protected]

Hi Ira,

Can you fix the subject to follow capitalization convention (use "git log --oneline") and say something more specific than "fix struct"?

Quoted 12 lines · On Mon, Nov 14, 2022 at 05:19:43PM -0800, [email protected] wrote:

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

Quoted 7 lines

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

I'm not really a fan of passing a parameter to say "this struct is on the stack" because that seems kind of error-prone and I don't know what the consequence of getting it wrong would be. Sounds like it *could* be some memory corruption or reading garbage data that would be hard to debug.

Do we have cases today where pci_doe_submit_task() can't do the kzalloc() as in your patch at [3]? If the current use cases allow a kzalloc(), why not do that now and defer this until it becomes an issue?

Bjorn

Quoted 2 lines

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

[3] https://lore.kernel.org/linux-cxl/Y2AnKB88ALYm9c5L@iweiny-desk3/

Quoted 115 lines
---
 drivers/cxl/core/pci.c  |  2 ++
 drivers/pci/doe.c       | 14 ++++++++++++--
 include/linux/pci-doe.h |  8 +++++---
 3 files changed, 19 insertions(+), 5 deletions(-)
drivers/cxl/core/pci.c +2 -0
index 9240df53ed87..a19c1fa0e2f4 100644
@@ -525,6 +525,7 @@ static int cxl_cdat_get_length(struct device *dev,
 	DECLARE_CDAT_DOE_TASK(CDAT_DOE_REQ(0), t);
 	int rc;
 
+	pci_doe_init_task(cdat_doe, &t.task, true);
 	rc = pci_doe_submit_task(cdat_doe, &t.task);
 	if (rc < 0) {
 		dev_err(dev, "DOE submit failed: %d", rc);
@@ -554,6 +555,7 @@ static int cxl_cdat_read_table(struct device *dev,
 		u32 *entry;
 		int rc;
 
+		pci_doe_init_task(cdat_doe, &t.task, true);
 		rc = pci_doe_submit_task(cdat_doe, &t.task);
 		if (rc < 0) {
 			dev_err(dev, "DOE submit failed: %d", rc);
drivers/pci/doe.c +12 -2
index e402f05068a5..cabeae4ae955 100644
@@ -319,6 +319,7 @@ static int pci_doe_discovery(struct pci_doe_mb *doe_mb, u8 *index, u16 *vid,
 	};
 	int rc;
 
+	pci_doe_init_task(doe_mb, &task, true);
 	rc = pci_doe_submit_task(doe_mb, &task);
 	if (rc < 0)
 		return rc;
@@ -495,6 +496,14 @@ bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type)
 }
 EXPORT_SYMBOL_GPL(pci_doe_supports_prot);
 
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack)
+{
+	task->doe_mb = doe_mb;
+	__INIT_WORK(&task->work, doe_statemachine_work, onstack);
+}
+EXPORT_SYMBOL_GPL(pci_doe_init_task);
+
 /**
  * pci_doe_submit_task() - Submit a task to be processed by the state machine
  *
@@ -517,6 +526,9 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (!pci_doe_supports_prot(doe_mb, task->prot.vid, task->prot.type))
 		return -EINVAL;
 
+	if (WARN_ON_ONCE(task->work.func != doe_statemachine_work))
+		return -EINVAL;
+
 	/*
 	 * DOE requests must be a whole number of DW and the response needs to
 	 * be big enough for at least 1 DW
@@ -528,8 +540,6 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (test_bit(PCI_DOE_FLAG_DEAD, &doe_mb->flags))
 		return -EIO;
 
-	task->doe_mb = doe_mb;
-	INIT_WORK(&task->work, doe_statemachine_work);
 	queue_work(doe_mb->work_queue, &task->work);
 	return 0;
 }
include/linux/pci-doe.h +5 -3
index ed9b4df792b8..457fc0e53d64 100644
@@ -31,8 +31,8 @@ struct pci_doe_mb;
  * @rv: Return value.  Length of received response or error (bytes)
  * @complete: Called when task is complete
  * @private: Private data for the consumer
- * @work: Used internally by the mailbox
- * @doe_mb: Used internally by the mailbox
+ * @work: Used internally by the mailbox [see pci_doe_init_task()]
+ * @doe_mb: Used internally by the mailbox [see pci_doe_init_task()]
  *
  * The payload sizes and rv are specified in bytes with the following
  * restrictions concerning the protocol.
@@ -53,7 +53,7 @@ struct pci_doe_task {
 	void (*complete)(struct pci_doe_task *task);
 	void *private;
 
-	/* No need for the user to initialize these fields */
+	/* Call pci_doe_init_task() for these */
 	struct work_struct work;
 	struct pci_doe_mb *doe_mb;
 };
@@ -72,6 +72,8 @@ struct pci_doe_task {
 
 struct pci_doe_mb *pcim_doe_create_mb(struct pci_dev *pdev, u16 cap_offset);
 bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type);
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack);
 int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task);
 
 #endif

base-commit: 30a0b95b1335e12efef89dd78518ed3e4a71a763

Signature · 1 line
2.37.2
author
#3 lore raw
Quoted 5 lines · Bjorn Helgaas

Hi Ira,

Can you fix the subject to follow capitalization convention (use "git log --oneline")

My apologies. I should have capitalized DOE.

Quoted 2 lines

and say something more specific than "fix struct"?

How about?

PCI/DOE: Fix initialization of work struct in pci_doe_task

Thanks for the review,
Ira

Quoted 155 lines · On Mon, Nov 14, 2022 at 05:19:43PM -0800, [email protected] wrote:
Quoted 12 lines

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

Quoted 7 lines

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

I'm not really a fan of passing a parameter to say "this struct is on the stack" because that seems kind of error-prone and I don't know what the consequence of getting it wrong would be. Sounds like it *could* be some memory corruption or reading garbage data that would be hard to debug.

Do we have cases today where pci_doe_submit_task() can't do the kzalloc() as in your patch at [3]? If the current use cases allow a kzalloc(), why not do that now and defer this until it becomes an issue?

Bjorn

Quoted 2 lines

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

[3] https://lore.kernel.org/linux-cxl/Y2AnKB88ALYm9c5L@iweiny-desk3/

Quoted 115 lines
---
 drivers/cxl/core/pci.c  |  2 ++
 drivers/pci/doe.c       | 14 ++++++++++++--
 include/linux/pci-doe.h |  8 +++++---
 3 files changed, 19 insertions(+), 5 deletions(-)
drivers/cxl/core/pci.c +2 -0
index 9240df53ed87..a19c1fa0e2f4 100644
@@ -525,6 +525,7 @@ static int cxl_cdat_get_length(struct device *dev,
 	DECLARE_CDAT_DOE_TASK(CDAT_DOE_REQ(0), t);
 	int rc;
 
+	pci_doe_init_task(cdat_doe, &t.task, true);
 	rc = pci_doe_submit_task(cdat_doe, &t.task);
 	if (rc < 0) {
 		dev_err(dev, "DOE submit failed: %d", rc);
@@ -554,6 +555,7 @@ static int cxl_cdat_read_table(struct device *dev,
 		u32 *entry;
 		int rc;
 
+		pci_doe_init_task(cdat_doe, &t.task, true);
 		rc = pci_doe_submit_task(cdat_doe, &t.task);
 		if (rc < 0) {
 			dev_err(dev, "DOE submit failed: %d", rc);
drivers/pci/doe.c +12 -2
index e402f05068a5..cabeae4ae955 100644
@@ -319,6 +319,7 @@ static int pci_doe_discovery(struct pci_doe_mb *doe_mb, u8 *index, u16 *vid,
 	};
 	int rc;
 
+	pci_doe_init_task(doe_mb, &task, true);
 	rc = pci_doe_submit_task(doe_mb, &task);
 	if (rc < 0)
 		return rc;
@@ -495,6 +496,14 @@ bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type)
 }
 EXPORT_SYMBOL_GPL(pci_doe_supports_prot);
 
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack)
+{
+	task->doe_mb = doe_mb;
+	__INIT_WORK(&task->work, doe_statemachine_work, onstack);
+}
+EXPORT_SYMBOL_GPL(pci_doe_init_task);
+
 /**
  * pci_doe_submit_task() - Submit a task to be processed by the state machine
  *
@@ -517,6 +526,9 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (!pci_doe_supports_prot(doe_mb, task->prot.vid, task->prot.type))
 		return -EINVAL;
 
+	if (WARN_ON_ONCE(task->work.func != doe_statemachine_work))
+		return -EINVAL;
+
 	/*
 	 * DOE requests must be a whole number of DW and the response needs to
 	 * be big enough for at least 1 DW
@@ -528,8 +540,6 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (test_bit(PCI_DOE_FLAG_DEAD, &doe_mb->flags))
 		return -EIO;
 
-	task->doe_mb = doe_mb;
-	INIT_WORK(&task->work, doe_statemachine_work);
 	queue_work(doe_mb->work_queue, &task->work);
 	return 0;
 }
include/linux/pci-doe.h +5 -3
index ed9b4df792b8..457fc0e53d64 100644
@@ -31,8 +31,8 @@ struct pci_doe_mb;
  * @rv: Return value.  Length of received response or error (bytes)
  * @complete: Called when task is complete
  * @private: Private data for the consumer
- * @work: Used internally by the mailbox
- * @doe_mb: Used internally by the mailbox
+ * @work: Used internally by the mailbox [see pci_doe_init_task()]
+ * @doe_mb: Used internally by the mailbox [see pci_doe_init_task()]
  *
  * The payload sizes and rv are specified in bytes with the following
  * restrictions concerning the protocol.
@@ -53,7 +53,7 @@ struct pci_doe_task {
 	void (*complete)(struct pci_doe_task *task);
 	void *private;
 
-	/* No need for the user to initialize these fields */
+	/* Call pci_doe_init_task() for these */
 	struct work_struct work;
 	struct pci_doe_mb *doe_mb;
 };
@@ -72,6 +72,8 @@ struct pci_doe_task {
 
 struct pci_doe_mb *pcim_doe_create_mb(struct pci_dev *pdev, u16 cap_offset);
 bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type);
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack);
 int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task);
 
 #endif

base-commit: 30a0b95b1335e12efef89dd78518ed3e4a71a763

Signature · 1 line
2.37.2
#4 lore raw
Quoted 6 lines · Ira Weiny
Quoted 1 line · Bjorn Helgaas

and say something more specific than "fix struct"?

How about?

PCI/DOE: Fix initialization of work struct in pci_doe_task

The importance of this has to do with whether something is on the stack, so I think something about that would be useful.

I'm afraid this subject line bike-shedding has made you overlook my other questions below ...

Quoted 154 lines
Quoted 154 lines · On Mon, Nov 14, 2022 at 05:19:43PM -0800, [email protected] wrote:
Quoted 12 lines

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

Quoted 7 lines

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

I'm not really a fan of passing a parameter to say "this struct is on the stack" because that seems kind of error-prone and I don't know what the consequence of getting it wrong would be. Sounds like it *could* be some memory corruption or reading garbage data that would be hard to debug.

Do we have cases today where pci_doe_submit_task() can't do the kzalloc() as in your patch at [3]? If the current use cases allow a kzalloc(), why not do that now and defer this until it becomes an issue?

Bjorn

Quoted 2 lines

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

[3] https://lore.kernel.org/linux-cxl/Y2AnKB88ALYm9c5L@iweiny-desk3/

Quoted 115 lines
---
 drivers/cxl/core/pci.c  |  2 ++
 drivers/pci/doe.c       | 14 ++++++++++++--
 include/linux/pci-doe.h |  8 +++++---
 3 files changed, 19 insertions(+), 5 deletions(-)
drivers/cxl/core/pci.c +2 -0
index 9240df53ed87..a19c1fa0e2f4 100644
@@ -525,6 +525,7 @@ static int cxl_cdat_get_length(struct device *dev,
 	DECLARE_CDAT_DOE_TASK(CDAT_DOE_REQ(0), t);
 	int rc;
 
+	pci_doe_init_task(cdat_doe, &t.task, true);
 	rc = pci_doe_submit_task(cdat_doe, &t.task);
 	if (rc < 0) {
 		dev_err(dev, "DOE submit failed: %d", rc);
@@ -554,6 +555,7 @@ static int cxl_cdat_read_table(struct device *dev,
 		u32 *entry;
 		int rc;
 
+		pci_doe_init_task(cdat_doe, &t.task, true);
 		rc = pci_doe_submit_task(cdat_doe, &t.task);
 		if (rc < 0) {
 			dev_err(dev, "DOE submit failed: %d", rc);
drivers/pci/doe.c +12 -2
index e402f05068a5..cabeae4ae955 100644
@@ -319,6 +319,7 @@ static int pci_doe_discovery(struct pci_doe_mb *doe_mb, u8 *index, u16 *vid,
 	};
 	int rc;
 
+	pci_doe_init_task(doe_mb, &task, true);
 	rc = pci_doe_submit_task(doe_mb, &task);
 	if (rc < 0)
 		return rc;
@@ -495,6 +496,14 @@ bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type)
 }
 EXPORT_SYMBOL_GPL(pci_doe_supports_prot);
 
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack)
+{
+	task->doe_mb = doe_mb;
+	__INIT_WORK(&task->work, doe_statemachine_work, onstack);
+}
+EXPORT_SYMBOL_GPL(pci_doe_init_task);
+
 /**
  * pci_doe_submit_task() - Submit a task to be processed by the state machine
  *
@@ -517,6 +526,9 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (!pci_doe_supports_prot(doe_mb, task->prot.vid, task->prot.type))
 		return -EINVAL;
 
+	if (WARN_ON_ONCE(task->work.func != doe_statemachine_work))
+		return -EINVAL;
+
 	/*
 	 * DOE requests must be a whole number of DW and the response needs to
 	 * be big enough for at least 1 DW
@@ -528,8 +540,6 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (test_bit(PCI_DOE_FLAG_DEAD, &doe_mb->flags))
 		return -EIO;
 
-	task->doe_mb = doe_mb;
-	INIT_WORK(&task->work, doe_statemachine_work);
 	queue_work(doe_mb->work_queue, &task->work);
 	return 0;
 }
include/linux/pci-doe.h +5 -3
index ed9b4df792b8..457fc0e53d64 100644
@@ -31,8 +31,8 @@ struct pci_doe_mb;
  * @rv: Return value.  Length of received response or error (bytes)
  * @complete: Called when task is complete
  * @private: Private data for the consumer
- * @work: Used internally by the mailbox
- * @doe_mb: Used internally by the mailbox
+ * @work: Used internally by the mailbox [see pci_doe_init_task()]
+ * @doe_mb: Used internally by the mailbox [see pci_doe_init_task()]
  *
  * The payload sizes and rv are specified in bytes with the following
  * restrictions concerning the protocol.
@@ -53,7 +53,7 @@ struct pci_doe_task {
 	void (*complete)(struct pci_doe_task *task);
 	void *private;
 
-	/* No need for the user to initialize these fields */
+	/* Call pci_doe_init_task() for these */
 	struct work_struct work;
 	struct pci_doe_mb *doe_mb;
 };
@@ -72,6 +72,8 @@ struct pci_doe_task {
 
 struct pci_doe_mb *pcim_doe_create_mb(struct pci_dev *pdev, u16 cap_offset);
 bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type);
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack);
 int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task);
 
 #endif

base-commit: 30a0b95b1335e12efef89dd78518ed3e4a71a763

Signature · 1 line
2.37.2
author
#5 lore raw
Quoted 6 lines · Bjorn Helgaas
Quoted 5 lines · Ira Weiny
Quoted 1 line · Bjorn Helgaas

and say something more specific than "fix struct"?

How about?

Quoted 7 lines
Quoted 1 line

PCI/DOE: Fix initialization of work struct in pci_doe_task

The importance of this has to do with whether something is on the stack, so I think something about that would be useful.

I'm afraid this subject line bike-shedding has made you overlook my other questions below ...

Oh... I see now.

Quoted 22 lines
Quoted 22 lines
Quoted 22 lines · On Mon, Nov 14, 2022 at 05:19:43PM -0800, [email protected] wrote:
Quoted 19 lines

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

I'm not really a fan of passing a parameter to say "this struct is on the stack" because that seems kind of error-prone and I don't know what the consequence of getting it wrong would be. Sounds like it *could* be some memory corruption or reading garbage data that would be hard to debug.

Do we have cases today where pci_doe_submit_task() can't do the kzalloc() as in your patch at [3]?

No.

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

If the current use cases allow a
kzalloc(), why not do that now and defer this until it becomes an issue?

I do like pci_doe_submit_task() handling this as an internal detail. I'm happy with that if you are.

I was just concerned about the restriction of context. Dan suggested this instead of passing a gfp parameter.

If you are happy with my original patch I will submit it instead. (With a better one liner.)

Thanks again for the review,
Ira

Quoted 123 lines
Quoted 123 lines
Quoted 123 lines

Bjorn

Quoted 2 lines

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

[3] https://lore.kernel.org/linux-cxl/Y2AnKB88ALYm9c5L@iweiny-desk3/

Quoted 115 lines
---
 drivers/cxl/core/pci.c  |  2 ++
 drivers/pci/doe.c       | 14 ++++++++++++--
 include/linux/pci-doe.h |  8 +++++---
 3 files changed, 19 insertions(+), 5 deletions(-)
drivers/cxl/core/pci.c +2 -0
index 9240df53ed87..a19c1fa0e2f4 100644
@@ -525,6 +525,7 @@ static int cxl_cdat_get_length(struct device *dev,
 	DECLARE_CDAT_DOE_TASK(CDAT_DOE_REQ(0), t);
 	int rc;
 
+	pci_doe_init_task(cdat_doe, &t.task, true);
 	rc = pci_doe_submit_task(cdat_doe, &t.task);
 	if (rc < 0) {
 		dev_err(dev, "DOE submit failed: %d", rc);
@@ -554,6 +555,7 @@ static int cxl_cdat_read_table(struct device *dev,
 		u32 *entry;
 		int rc;
 
+		pci_doe_init_task(cdat_doe, &t.task, true);
 		rc = pci_doe_submit_task(cdat_doe, &t.task);
 		if (rc < 0) {
 			dev_err(dev, "DOE submit failed: %d", rc);
drivers/pci/doe.c +12 -2
index e402f05068a5..cabeae4ae955 100644
@@ -319,6 +319,7 @@ static int pci_doe_discovery(struct pci_doe_mb *doe_mb, u8 *index, u16 *vid,
 	};
 	int rc;
 
+	pci_doe_init_task(doe_mb, &task, true);
 	rc = pci_doe_submit_task(doe_mb, &task);
 	if (rc < 0)
 		return rc;
@@ -495,6 +496,14 @@ bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type)
 }
 EXPORT_SYMBOL_GPL(pci_doe_supports_prot);
 
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack)
+{
+	task->doe_mb = doe_mb;
+	__INIT_WORK(&task->work, doe_statemachine_work, onstack);
+}
+EXPORT_SYMBOL_GPL(pci_doe_init_task);
+
 /**
  * pci_doe_submit_task() - Submit a task to be processed by the state machine
  *
@@ -517,6 +526,9 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (!pci_doe_supports_prot(doe_mb, task->prot.vid, task->prot.type))
 		return -EINVAL;
 
+	if (WARN_ON_ONCE(task->work.func != doe_statemachine_work))
+		return -EINVAL;
+
 	/*
 	 * DOE requests must be a whole number of DW and the response needs to
 	 * be big enough for at least 1 DW
@@ -528,8 +540,6 @@ int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task)
 	if (test_bit(PCI_DOE_FLAG_DEAD, &doe_mb->flags))
 		return -EIO;
 
-	task->doe_mb = doe_mb;
-	INIT_WORK(&task->work, doe_statemachine_work);
 	queue_work(doe_mb->work_queue, &task->work);
 	return 0;
 }
include/linux/pci-doe.h +5 -3
index ed9b4df792b8..457fc0e53d64 100644
@@ -31,8 +31,8 @@ struct pci_doe_mb;
  * @rv: Return value.  Length of received response or error (bytes)
  * @complete: Called when task is complete
  * @private: Private data for the consumer
- * @work: Used internally by the mailbox
- * @doe_mb: Used internally by the mailbox
+ * @work: Used internally by the mailbox [see pci_doe_init_task()]
+ * @doe_mb: Used internally by the mailbox [see pci_doe_init_task()]
  *
  * The payload sizes and rv are specified in bytes with the following
  * restrictions concerning the protocol.
@@ -53,7 +53,7 @@ struct pci_doe_task {
 	void (*complete)(struct pci_doe_task *task);
 	void *private;
 
-	/* No need for the user to initialize these fields */
+	/* Call pci_doe_init_task() for these */
 	struct work_struct work;
 	struct pci_doe_mb *doe_mb;
 };
@@ -72,6 +72,8 @@ struct pci_doe_task {
 
 struct pci_doe_mb *pcim_doe_create_mb(struct pci_dev *pdev, u16 cap_offset);
 bool pci_doe_supports_prot(struct pci_doe_mb *doe_mb, u16 vid, u8 type);
+void pci_doe_init_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task,
+		       bool onstack);
 int pci_doe_submit_task(struct pci_doe_mb *doe_mb, struct pci_doe_task *task);
 
 #endif

base-commit: 30a0b95b1335e12efef89dd78518ed3e4a71a763

Signature · 1 line
2.37.2
#6 lore raw
Quoted 42 lines · Ira Weiny
Quoted 26 lines · Bjorn Helgaas
Quoted 25 lines · Ira Weiny
Quoted 24 lines · Bjorn Helgaas
Quoted 22 lines · On Mon, Nov 14, 2022 at 05:19:43PM -0800, [email protected] wrote:

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

I'm not really a fan of passing a parameter to say "this struct is on the stack" because that seems kind of error-prone and I don't know what the consequence of getting it wrong would be. Sounds like it *could* be some memory corruption or reading garbage data that would be hard to debug.

Do we have cases today where pci_doe_submit_task() can't do the kzalloc() as in your patch at [3]?

No.

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

If the current use cases allow a
kzalloc(), why not do that now and defer this until it becomes an issue?

Quoted 8 lines

I do like pci_doe_submit_task() handling this as an internal detail. I'm happy with that if you are.

I was just concerned about the restriction of context. Dan suggested this instead of passing a gfp parameter.

If you are happy with my original patch I will submit it instead. (With a better one liner.)

I don't know what's coming as far as pci_doe_submit_task() callers. If there's some imminent caller that will require atomic context, I guess we could solve it now. But DOE doesn't really seem like an atomic context thing to begin with, so maybe we could postpone dealing with it.

That patch in [3] is more complicated than I expected, but I admit I haven't looked closely.

Bjorn

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 2 lines

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

[3] https://lore.kernel.org/linux-cxl/Y2AnKB88ALYm9c5L@iweiny-desk3/

#7 lore raw
In reply to [email protected]
Quoted 16 lines · On Mon, Nov 14, 2022 at 05:19:43PM -0800, [email protected] wrote:

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Quoted 6 lines

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

We have object_is_on_stack(), included from <linux/sched/task_stack.h>.

So you could just autosense in pci_doe_submit_task() whether pci_doe_task is on the stack and call the appropriate INIT_WORK variant.

Any reason not to do that?

Thanks,

Lukas

#8 lore raw
Quoted 21 lines · Lukas Wunner
Quoted 20 lines · On Mon, Nov 14, 2022 at 05:19:43PM -0800, [email protected] wrote:

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

Quoted 8 lines
Quoted 2 lines

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

We have object_is_on_stack(), included from <linux/sched/task_stack.h>.

So you could just autosense in pci_doe_submit_task() whether pci_doe_task is on the stack and call the appropriate INIT_WORK variant.

Nifty, I had no idea object_is_on_stack() existed, thank you!

I wonder if there's an opportunity to use object_is_on_stack() somewhere in the INIT_WORK() path to find usage mistakes.

Adding it in pci_doe_submit_task() would add some complexity, so I'm not sure whether it's worth adding it unless we actually have uses for both cases.

Bjorn

author
#9 lore raw
Quoted 25 lines · Bjorn Helgaas
Quoted 24 lines · Lukas Wunner
Quoted 22 lines · On Mon, Nov 14, 2022 at 05:19:43PM -0800, [email protected] wrote:

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

Quoted 7 lines
Quoted 5 lines

We have object_is_on_stack(), included from <linux/sched/task_stack.h>.

So you could just autosense in pci_doe_submit_task() whether pci_doe_task is on the stack and call the appropriate INIT_WORK variant.

Nifty, I had no idea object_is_on_stack() existed, thank you!

Indeed! Neither did I! thanks!

Quoted 3 lines

I wonder if there's an opportunity to use object_is_on_stack() somewhere in the INIT_WORK() path to find usage mistakes.

I'm thinking we could make INIT_WORK do the right thing all the time. Not sure what the overhead of object_is_on_stack() is.

Quoted 4 lines

Adding it in pci_doe_submit_task() would add some complexity, so I'm not sure whether it's worth adding it unless we actually have uses for both cases.

I think if we don't do something we have to document that pci_doe_submit_task() only works with tasks on the stack.

I would rather just make pci_doe_submit_task() correct and not complicate the callers. object_is_on_stack() can't be enough overhead to be worried about in this call path can it?

Actually after writing all that I wonder if we can't push the use of object_is_on_stack() into the debug code? Something like below (completely untested)? I think this could be pushed even further down but I'd like to get opinions before attempting a change which will have a wider blast radius.

Ira

include/linux/workqueue.h +8 -8
index a0143dd24430..4cc50b554a29 100644
@@ -199,7 +199,7 @@ struct execute_work {
        struct delayed_work n = __DELAYED_WORK_INITIALIZER(n, f, TIMER_DEFERRABLE)
 
 #ifdef CONFIG_DEBUG_OBJECTS_WORK
-extern void __init_work(struct work_struct *work, int onstack);
+extern void __init_work(struct work_struct *work);
 extern void destroy_work_on_stack(struct work_struct *work);
 extern void destroy_delayed_work_on_stack(struct delayed_work *work);
 static inline unsigned int work_static(struct work_struct *work)
@@ -207,7 +207,7 @@ static inline unsigned int work_static(struct work_struct *work)
        return *work_data_bits(work) & WORK_STRUCT_STATIC;
 }
 #else
-static inline void __init_work(struct work_struct *work, int onstack) { }
+static inline void __init_work(struct work_struct *work) { }
 static inline void destroy_work_on_stack(struct work_struct *work) { }
 static inline void destroy_delayed_work_on_stack(struct delayed_work *work) { }
 static inline unsigned int work_static(struct work_struct *work) { return 0; }
@@ -221,20 +221,20 @@ static inline unsigned int work_static(struct work_struct *work) { return 0; }
  * to generate better code.
  */
 #ifdef CONFIG_LOCKDEP
-#define __INIT_WORK(_work, _func, _onstack)                            \
+#define __INIT_WORK(_work, _func)                                      \
        do {                                                            \
                static struct lock_class_key __key;                     \
                                                                        \
-               __init_work((_work), _onstack);                         \
+               __init_work(_work);                                     \
                (_work)->data = (atomic_long_t) WORK_DATA_INIT();       \
                lockdep_init_map(&(_work)->lockdep_map, "(work_completion)"#_work, &__key, 0); \
                INIT_LIST_HEAD(&(_work)->entry);                        \
                (_work)->func = (_func);                                \
        } while (0)
 #else
-#define __INIT_WORK(_work, _func, _onstack)                            \
+#define __INIT_WORK(_work, _func)                                      \
        do {                                                            \
-               __init_work((_work), _onstack);                         \
+               __init_work(_work);                                     \
                (_work)->data = (atomic_long_t) WORK_DATA_INIT();       \
                INIT_LIST_HEAD(&(_work)->entry);                        \
                (_work)->func = (_func);                                \
@@ -242,10 +242,10 @@ static inline unsigned int work_static(struct work_struct *work) { return 0; }
 #endif
 
 #define INIT_WORK(_work, _func)                                                \
-       __INIT_WORK((_work), (_func), 0)
+       __INIT_WORK((_work), (_func))
 
 #define INIT_WORK_ONSTACK(_work, _func)                                        \
-       __INIT_WORK((_work), (_func), 1)
+       __INIT_WORK((_work), (_func))
 
 #define __INIT_DELAYED_WORK(_work, _func, _tflags)                     \
        do {                                                            \
kernel/workqueue.c +2 -2
index 7cd5f5e7e0a1..7d87300cfbc6 100644
@@ -515,9 +515,9 @@ static inline void debug_work_deactivate(struct work_struct *work)
        debug_object_deactivate(work, &work_debug_descr);
 }
 
-void __init_work(struct work_struct *work, int onstack)
+void __init_work(struct work_struct *work)
 {
-       if (onstack)
+       if (object_is_on_stack(work))
                debug_object_init_on_stack(work, &work_debug_descr);
        else
                debug_object_init(work, &work_debug_descr);
#10 lore raw
Quoted 51 lines · Ira Weiny
Quoted 25 lines · Bjorn Helgaas
Quoted 24 lines · Lukas Wunner
Quoted 22 lines · On Mon, Nov 14, 2022 at 05:19:43PM -0800, [email protected] wrote:

From: Ira Weiny <[email protected]>

The callers of pci_doe_submit_task() allocate the pci_doe_task on the stack. This causes the work structure to be allocated on the stack without pci_doe_submit_task() knowing. Work item initialization needs to be done with either INIT_WORK_ONSTACK() or INIT_WORK() depending on how the work item is allocated.

Jonathan suggested creating doe task allocation macros such as DECLARE_CDAT_DOE_TASK_ONSTACK().[1] The issue with this is the work function is not known to the callers and must be initialized correctly.

A follow up suggestion was to have an internal 'pci_doe_work' item allocated by pci_doe_submit_task().[2] This requires an allocation which could restrict the context where tasks are used.

Compromise with an intermediate step to initialize the task struct with a new call pci_doe_init_task() which must be called prior to submit task.

[1] https://lore.kernel.org/linux-cxl/[email protected]/T/#m88a7f50dcce52f30c8bf5c3dcc06fa9843b54a2d
[2] https://lore.kernel.org/linux-cxl/[email protected]/T/#m63c636c5135f304480370924f4d03c00357be667

Quoted 7 lines
Quoted 5 lines

We have object_is_on_stack(), included from <linux/sched/task_stack.h>.

So you could just autosense in pci_doe_submit_task() whether pci_doe_task is on the stack and call the appropriate INIT_WORK variant.

Nifty, I had no idea object_is_on_stack() existed, thank you!

Indeed! Neither did I! thanks!

Quoted 3 lines

I wonder if there's an opportunity to use object_is_on_stack() somewhere in the INIT_WORK() path to find usage mistakes.

I'm thinking we could make INIT_WORK do the right thing all the time. Not sure what the overhead of object_is_on_stack() is.

Quoted 4 lines

Adding it in pci_doe_submit_task() would add some complexity, so I'm not sure whether it's worth adding it unless we actually have uses for both cases.

I think if we don't do something we have to document that pci_doe_submit_task() only works with tasks on the stack.

Quoted 8 lines

I would rather just make pci_doe_submit_task() correct and not complicate the callers. object_is_on_stack() can't be enough overhead to be worried about in this call path can it?

Actually after writing all that I wonder if we can't push the use of object_is_on_stack() into the debug code? Something like below (completely untested)? I think this could be pushed even further down but I'd like to get opinions before attempting a change which will have a wider blast radius.

This looks reasonable, but I would do it after and independently of introducing the autosensing version of pci_doe_submit_task(). Then you can pursue this line of thinking and come back to simplify pci_doe_submit_task() if it indeed moves forward.