Skip to contentSkip to conversation

remoteproc: bug fixes

PATCH In review Alex Elder 4 participants 9 messages Apr 3, 2020 – Apr 17, 2020

Subsystems (2)
lore
Share & tools
lore

Alex Elder

author
#0 lore raw

cover remoteproc: bug fixes

This series fixes some bugs in remoteproc error paths. The last patch is derived from something I posted before, but has been updated to based on Linus' current master branch.

-Alex

Alex Elder (3):
remoteproc: fix a bug in rproc_alloc()
remoteproc: qcom_q6v5_mss: fix a bug in q6v5_probe()
remoteproc: qcom_q6v5_mss: fix q6v5_probe() error paths

 drivers/remoteproc/qcom_q6v5_mss.c   | 33 +++++++++++++++++-----------
 drivers/remoteproc/remoteproc_core.c |  4 ++--
 2 files changed, 22 insertions(+), 15 deletions(-)
Signature · 1 line
2.20.1
author
#1 lore raw

PATCH 1/3 remoteproc: fix a bug in rproc_alloc()

If ida_simple_get() returns an error when called in rproc_alloc(), put_device() is called to clean things up. By this time the rproc device type has been assigned, with rproc_type_release() as the release function.

The first thing rproc_type_release() does is call:

    idr_destroy(&rproc->notifyids);

But at the time the ida_simple_get() call is made, the notifyids field in the remoteproc structure has not been initialized.

I'm not actually sure this case causes an observable problem, but it's incorrect. Fix this by initializing the notifyids field before calling ida_simple_get() in rproc_alloc().

---
 drivers/remoteproc/remoteproc_core.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/remoteproc/remoteproc_core.c +2 -2
index e12a54e67588..59b6eb22f01c 100644
@@ -2054,6 +2054,8 @@ struct rproc *rproc_alloc(struct device *dev, const char *name,
 	rproc->dev.class = &rproc_class;
 	rproc->dev.driver_data = rproc;
 
+	idr_init(&rproc->notifyids);
+
 	/* Assign a unique device index and name */
 	rproc->index = ida_simple_get(&rproc_dev_index, 0, 0, GFP_KERNEL);
 	if (rproc->index < 0) {
@@ -2078,8 +2080,6 @@ struct rproc *rproc_alloc(struct device *dev, const char *name,
 
 	mutex_init(&rproc->lock);
 
-	idr_init(&rproc->notifyids);
-
 	INIT_LIST_HEAD(&rproc->carveouts);
 	INIT_LIST_HEAD(&rproc->mappings);
 	INIT_LIST_HEAD(&rproc->traces);
Signature · 1 line
2.20.1
maintainer
#2 lore raw

Re: [PATCH 1/3] remoteproc: fix a bug in rproc_alloc()

Hi Alex,

Quoted 7 lines · Alex Elder

If ida_simple_get() returns an error when called in rproc_alloc(), put_device() is called to clean things up. By this time the rproc device type has been assigned, with rproc_type_release() as the release function.

The first thing rproc_type_release() does is call:

Quoted 9 lines
    idr_destroy(&rproc->notifyids);

But at the time the ida_simple_get() call is made, the notifyids field in the remoteproc structure has not been initialized.

I'm not actually sure this case causes an observable problem, but it's incorrect. Fix this by initializing the notifyids field before calling ida_simple_get() in rproc_alloc().

Both Suman and I are meddling in function rproc_alloc() for our respective work [1][2]. I will add this patch to a set that refactors rproc_alloc() as soon as v5.7-rc1 comes out. That way we can all base our work on the same foundation and Bjorn doesn't have to fix 3 different merge conflicts.

Thanks,
Mathieu

[1]. https://patchwork.kernel.org/patch/11456385/
[2]. https://patchwork.kernel.org/project/linux-remoteproc/list/?series=261069

Quoted 30 lines
---
 drivers/remoteproc/remoteproc_core.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/remoteproc/remoteproc_core.c +2 -3
index e12a54e67588..59b6eb22f01c 100644
@@ -2054,6 +2054,8 @@ struct rproc *rproc_alloc(struct device *dev, const char *name,
        rproc->dev.class = &rproc_class;
        rproc->dev.driver_data = rproc;
 
+       idr_init(&rproc->notifyids);
+
        /* Assign a unique device index and name */
        rproc->index = ida_simple_get(&rproc_dev_index, 0, 0, GFP_KERNEL);
        if (rproc->index < 0) {
@@ -2078,8 +2080,6 @@ struct rproc *rproc_alloc(struct device *dev, const char *name,
 
        mutex_init(&rproc->lock);
 
-       idr_init(&rproc->notifyids);
-
        INIT_LIST_HEAD(&rproc->carveouts);
        INIT_LIST_HEAD(&rproc->mappings);
        INIT_LIST_HEAD(&rproc->traces);
--

2.20.1

author
#3 lore raw

Re: [PATCH 1/3] remoteproc: fix a bug in rproc_alloc()

Quoted 20 lines · Mathieu Poirier

Hi Alex,

Quoted 7 lines · Alex Elder

If ida_simple_get() returns an error when called in rproc_alloc(), put_device() is called to clean things up. By this time the rproc device type has been assigned, with rproc_type_release() as the release function.

The first thing rproc_type_release() does is call:

Quoted 9 lines
    idr_destroy(&rproc->notifyids);

But at the time the ida_simple_get() call is made, the notifyids field in the remoteproc structure has not been initialized.

I'm not actually sure this case causes an observable problem, but it's incorrect. Fix this by initializing the notifyids field before calling ida_simple_get() in rproc_alloc().

Quoted 8 lines

Both Suman and I are meddling in function rproc_alloc() for our respective work [1][2]. I will add this patch to a set that refactors rproc_alloc() as soon as v5.7-rc1 comes out. That way we can all base our work on the same foundation and Bjorn doesn't have to fix 3 different merge conflicts.

Thanks,
Mathieu

Fine with me. Thanks a lot. -Alex

Quoted 33 lines

[1]. https://patchwork.kernel.org/patch/11456385/
[2]. https://patchwork.kernel.org/project/linux-remoteproc/list/?series=261069

Quoted 30 lines
---
 drivers/remoteproc/remoteproc_core.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/remoteproc/remoteproc_core.c +2 -3
index e12a54e67588..59b6eb22f01c 100644
@@ -2054,6 +2054,8 @@ struct rproc *rproc_alloc(struct device *dev, const char *name,
        rproc->dev.class = &rproc_class;
        rproc->dev.driver_data = rproc;
 
+       idr_init(&rproc->notifyids);
+
        /* Assign a unique device index and name */
        rproc->index = ida_simple_get(&rproc_dev_index, 0, 0, GFP_KERNEL);
        if (rproc->index < 0) {
@@ -2078,8 +2080,6 @@ struct rproc *rproc_alloc(struct device *dev, const char *name,
 
        mutex_init(&rproc->lock);
 
-       idr_init(&rproc->notifyids);
-
        INIT_LIST_HEAD(&rproc->carveouts);
        INIT_LIST_HEAD(&rproc->mappings);
        INIT_LIST_HEAD(&rproc->traces);
--

2.20.1

author
#4 lore raw

PATCH 2/3 remoteproc: qcom_q6v5_mss: fix a bug in q6v5_probe()

In reply to Alex Elder

If looking up the DT "firmware-name" property fails in q6v6_probe(), the function returns without freeing the remoteproc structure that has been allocated. Fix this by jumping to the free_rproc label, which takes care of this.

---
 drivers/remoteproc/qcom_q6v5_mss.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/remoteproc/qcom_q6v5_mss.c +1 -1
index ce49c3236ff7..60cdf699ea80 100644
@@ -1675,7 +1675,7 @@ static int q6v5_probe(struct platform_device *pdev)
 	ret = of_property_read_string_index(pdev->dev.of_node, "firmware-name",
 					    1, &qproc->hexagon_mdt_image);
 	if (ret < 0 && ret != -EINVAL)
-		return ret;
+		goto free_rproc;
 
 	platform_set_drvdata(pdev, qproc);
 
Signature · 1 line
2.20.1
#5 lore raw

Re: [PATCH 2/3] remoteproc: qcom_q6v5_mss: fix a bug in q6v5_probe()

Quoted 5 lines · Alex Elder

If looking up the DT "firmware-name" property fails in q6v6_probe(), the function returns without freeing the remoteproc structure that has been allocated. Fix this by jumping to the free_rproc label, which takes care of this.

Please add the Fixes: line.

Quoted 19 lines
---
  drivers/remoteproc/qcom_q6v5_mss.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/remoteproc/qcom_q6v5_mss.c +1 -1
index ce49c3236ff7..60cdf699ea80 100644
@@ -1675,7 +1675,7 @@ static int q6v5_probe(struct platform_device *pdev)
  	ret = of_property_read_string_index(pdev->dev.of_node, "firmware-name",
  					    1, &qproc->hexagon_mdt_image);
  	if (ret < 0 && ret != -EINVAL)
-		return ret;
+		goto free_rproc;
 
  	platform_set_drvdata(pdev, qproc);
 
author
#6 lore raw

PATCH 3/3 remoteproc: qcom_q6v5_mss: fix q6v5_probe() error paths

In reply to Alex Elder

If an error occurs in q6v5_probe() after the proxy power domains are attached, but before qcom_add_ipa_notify_subdev() is called, qcom_remove_ipa_notify_subdev() is called in the error path, which is a bug. Fix this by having that call be reached through a different label.

Additionally, if qcom_add_sysmon_subdev() returns an error, the subdevs that had already been added will not be properly removed. Fix this by having the added subdevs (including the IPA notify one) be removed in this case.

Finally, arrange for the sysmon subdev to be removed before the rest in the event rproc_add() returns an error.

Have cleanup activity done in q6v5_remove() be done in the reverse order they are set up in q6v5_probe() (the same order they're done in the q6v5_probe() error path). Use a local variable for the remoteproc pointer, which is used repeatedly.

Remove errant semicolons at the end of two function blocks.

---
 drivers/remoteproc/qcom_q6v5_mss.c | 31 ++++++++++++++++++------------
 1 file changed, 19 insertions(+), 12 deletions(-)
drivers/remoteproc/qcom_q6v5_mss.c +19 -12
index 60cdf699ea80..5475d4f808a8 100644
@@ -367,7 +367,7 @@ static int q6v5_pds_enable(struct q6v5 *qproc, struct device **pds,
 	}
 
 	return ret;
-};
+}
 
 static void q6v5_pds_disable(struct q6v5 *qproc, struct device **pds,
 			     size_t pd_count)
@@ -1527,7 +1527,7 @@ static int q6v5_pds_attach(struct device *dev, struct device **devs,
 		dev_pm_domain_detach(devs[i], false);
 
 	return ret;
-};
+}
 
 static void q6v5_pds_detach(struct q6v5 *qproc, struct device **pds,
 			    size_t pd_count)
@@ -1766,17 +1766,23 @@ static int q6v5_probe(struct platform_device *pdev)
 	qproc->sysmon = qcom_add_sysmon_subdev(rproc, "modem", 0x12);
 	if (IS_ERR(qproc->sysmon)) {
 		ret = PTR_ERR(qproc->sysmon);
-		goto detach_proxy_pds;
+		goto remove_subdevs;
 	}
 
 	ret = rproc_add(rproc);
 	if (ret)
-		goto detach_proxy_pds;
+		goto remove_sysmon_subdev;
 
 	return 0;
 
+remove_sysmon_subdev:
+	qcom_remove_sysmon_subdev(qproc->sysmon);
+remove_subdevs:
+	qcom_remove_ipa_notify_subdev(qproc->rproc, &qproc->ipa_notify_subdev);
+	qcom_remove_ssr_subdev(rproc, &qproc->ssr_subdev);
+	qcom_remove_smd_subdev(rproc, &qproc->smd_subdev);
+	qcom_remove_glink_subdev(rproc, &qproc->glink_subdev);
 detach_proxy_pds:
-	qcom_remove_ipa_notify_subdev(qproc->rproc, &qproc->ipa_notify_subdev);
 	q6v5_pds_detach(qproc, qproc->proxy_pds, qproc->proxy_pd_count);
 detach_active_pds:
 	q6v5_pds_detach(qproc, qproc->active_pds, qproc->active_pd_count);
@@ -1789,19 +1795,20 @@ static int q6v5_probe(struct platform_device *pdev)
 static int q6v5_remove(struct platform_device *pdev)
 {
 	struct q6v5 *qproc = platform_get_drvdata(pdev);
+	struct rproc *rproc = qproc->rproc;
 
-	rproc_del(qproc->rproc);
+	rproc_del(rproc);
 
 	qcom_remove_sysmon_subdev(qproc->sysmon);
-	qcom_remove_ipa_notify_subdev(qproc->rproc, &qproc->ipa_notify_subdev);
-	qcom_remove_glink_subdev(qproc->rproc, &qproc->glink_subdev);
-	qcom_remove_smd_subdev(qproc->rproc, &qproc->smd_subdev);
-	qcom_remove_ssr_subdev(qproc->rproc, &qproc->ssr_subdev);
+	qcom_remove_ipa_notify_subdev(rproc, &qproc->ipa_notify_subdev);
+	qcom_remove_ssr_subdev(rproc, &qproc->ssr_subdev);
+	qcom_remove_smd_subdev(rproc, &qproc->smd_subdev);
+	qcom_remove_glink_subdev(rproc, &qproc->glink_subdev);
 
-	q6v5_pds_detach(qproc, qproc->active_pds, qproc->active_pd_count);
 	q6v5_pds_detach(qproc, qproc->proxy_pds, qproc->proxy_pd_count);
+	q6v5_pds_detach(qproc, qproc->active_pds, qproc->active_pd_count);
 
-	rproc_free(qproc->rproc);
+	rproc_free(rproc);
 
 	return 0;
 }
Signature · 1 line
2.20.1
#7 lore raw

Re: [PATCH 3/3] remoteproc: qcom_q6v5_mss: fix q6v5_probe() error paths

Quoted 14 lines · Alex Elder

If an error occurs in q6v5_probe() after the proxy power domains are attached, but before qcom_add_ipa_notify_subdev() is called, qcom_remove_ipa_notify_subdev() is called in the error path, which is a bug. Fix this by having that call be reached through a different label.

Additionally, if qcom_add_sysmon_subdev() returns an error, the subdevs that had already been added will not be properly removed. Fix this by having the added subdevs (including the IPA notify one) be removed in this case.

Finally, arrange for the sysmon subdev to be removed before the rest in the event rproc_add() returns an error.

Quoted 7 lines

Have cleanup activity done in q6v5_remove() be done in the reverse order they are set up in q6v5_probe() (the same order they're done in the q6v5_probe() error path). Use a local variable for the remoteproc pointer, which is used repeatedly.

Remove errant semicolons at the end of two function blocks.

Same as Patch 2, Fixes: line would be good.

regards
Suman

Quoted 83 lines
---
  drivers/remoteproc/qcom_q6v5_mss.c | 31 ++++++++++++++++++------------
  1 file changed, 19 insertions(+), 12 deletions(-)
drivers/remoteproc/qcom_q6v5_mss.c +19 -12
index 60cdf699ea80..5475d4f808a8 100644
@@ -367,7 +367,7 @@ static int q6v5_pds_enable(struct q6v5 *qproc, struct device **pds,
  	}
 
  	return ret;
-};
+}
 
  static void q6v5_pds_disable(struct q6v5 *qproc, struct device **pds,
  			     size_t pd_count)
@@ -1527,7 +1527,7 @@ static int q6v5_pds_attach(struct device *dev, struct device **devs,
  		dev_pm_domain_detach(devs[i], false);
 
  	return ret;
-};
+}
 
  static void q6v5_pds_detach(struct q6v5 *qproc, struct device **pds,
  			    size_t pd_count)
@@ -1766,17 +1766,23 @@ static int q6v5_probe(struct platform_device *pdev)
  	qproc->sysmon = qcom_add_sysmon_subdev(rproc, "modem", 0x12);
  	if (IS_ERR(qproc->sysmon)) {
  		ret = PTR_ERR(qproc->sysmon);
-		goto detach_proxy_pds;
+		goto remove_subdevs;
  	}
 
  	ret = rproc_add(rproc);
  	if (ret)
-		goto detach_proxy_pds;
+		goto remove_sysmon_subdev;
 
  	return 0;
 
+remove_sysmon_subdev:
+	qcom_remove_sysmon_subdev(qproc->sysmon);
+remove_subdevs:
+	qcom_remove_ipa_notify_subdev(qproc->rproc, &qproc->ipa_notify_subdev);
+	qcom_remove_ssr_subdev(rproc, &qproc->ssr_subdev);
+	qcom_remove_smd_subdev(rproc, &qproc->smd_subdev);
+	qcom_remove_glink_subdev(rproc, &qproc->glink_subdev);
  detach_proxy_pds:
-	qcom_remove_ipa_notify_subdev(qproc->rproc, &qproc->ipa_notify_subdev);
  	q6v5_pds_detach(qproc, qproc->proxy_pds, qproc->proxy_pd_count);
  detach_active_pds:
  	q6v5_pds_detach(qproc, qproc->active_pds, qproc->active_pd_count);
@@ -1789,19 +1795,20 @@ static int q6v5_probe(struct platform_device *pdev)
  static int q6v5_remove(struct platform_device *pdev)
  {
  	struct q6v5 *qproc = platform_get_drvdata(pdev);
+	struct rproc *rproc = qproc->rproc;
 
-	rproc_del(qproc->rproc);
+	rproc_del(rproc);
 
  	qcom_remove_sysmon_subdev(qproc->sysmon);
-	qcom_remove_ipa_notify_subdev(qproc->rproc, &qproc->ipa_notify_subdev);
-	qcom_remove_glink_subdev(qproc->rproc, &qproc->glink_subdev);
-	qcom_remove_smd_subdev(qproc->rproc, &qproc->smd_subdev);
-	qcom_remove_ssr_subdev(qproc->rproc, &qproc->ssr_subdev);
+	qcom_remove_ipa_notify_subdev(rproc, &qproc->ipa_notify_subdev);
+	qcom_remove_ssr_subdev(rproc, &qproc->ssr_subdev);
+	qcom_remove_smd_subdev(rproc, &qproc->smd_subdev);
+	qcom_remove_glink_subdev(rproc, &qproc->glink_subdev);
 
-	q6v5_pds_detach(qproc, qproc->active_pds, qproc->active_pd_count);
  	q6v5_pds_detach(qproc, qproc->proxy_pds, qproc->proxy_pd_count);
+	q6v5_pds_detach(qproc, qproc->active_pds, qproc->active_pd_count);
 
-	rproc_free(qproc->rproc);
+	rproc_free(rproc);
 
  	return 0;
  }
Tested-by on the v1 series
#8 lore raw
In reply to Alex Elder
Quoted 5 lines · Alex Elder

This series fixes some bugs in remoteproc error paths. The last patch is derived from something I posted before, but has been updated to based on Linus' current master branch.

Oof. Apparently I didn't hit reply all but only sent my tested-by to just Alex (now 10 days ago)

This set resolves regressions we've seen with 5.7-rc1 on db845c, and it would be really nice to see them land in 5.7-rc2.

thanks
-john