* [PATCH 1/3] misc: pci_endpoint_test: Fix failure path return values in probe
2017-10-11 8:44 [PATCH 0/3] pci_endpoint_test: Fix module removal/insertion cycle Kishon Vijay Abraham I
@ 2017-10-11 8:44 ` Kishon Vijay Abraham I
2017-10-11 8:44 ` [PATCH 2/3] misc: pci_endpoint_test: Fix pci_endpoint_test not releasing resources on remove Kishon Vijay Abraham I
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: Kishon Vijay Abraham I @ 2017-10-11 8:44 UTC (permalink / raw)
To: Arnd Bergmann, Greg Kroah-Hartman, Kishon Vijay Abraham I
Cc: linux-pci, linux-omap, linux-kernel, nsekhar
Return value of pci_endpoint_test_probe is not set properly in
a couple of failure cases. Fix it here.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
---
drivers/misc/pci_endpoint_test.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c
index deb203026496..e787a63a321a 100644
--- a/drivers/misc/pci_endpoint_test.c
+++ b/drivers/misc/pci_endpoint_test.c
@@ -533,6 +533,7 @@ static int pci_endpoint_test_probe(struct pci_dev *pdev,
test->base = test->bar[test_reg_bar];
if (!test->base) {
+ err = -ENOMEM;
dev_err(dev, "Cannot perform PCI test without BAR%d\n",
test_reg_bar);
goto err_iounmap;
@@ -542,6 +543,7 @@ static int pci_endpoint_test_probe(struct pci_dev *pdev,
id = ida_simple_get(&pci_endpoint_test_ida, 0, 0, GFP_KERNEL);
if (id < 0) {
+ err = id;
dev_err(dev, "unable to get id\n");
goto err_iounmap;
}
--
2.11.0
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH 2/3] misc: pci_endpoint_test: Fix pci_endpoint_test not releasing resources on remove
2017-10-11 8:44 [PATCH 0/3] pci_endpoint_test: Fix module removal/insertion cycle Kishon Vijay Abraham I
2017-10-11 8:44 ` [PATCH 1/3] misc: pci_endpoint_test: Fix failure path return values in probe Kishon Vijay Abraham I
@ 2017-10-11 8:44 ` Kishon Vijay Abraham I
2017-10-11 8:44 ` [PATCH 3/3] misc: pci_endpoint_test: Fix BUG_ON error during pci_disable_msi Kishon Vijay Abraham I
2017-10-17 19:16 ` [PATCH 0/3] pci_endpoint_test: Fix module removal/insertion cycle Bjorn Helgaas
3 siblings, 0 replies; 5+ messages in thread
From: Kishon Vijay Abraham I @ 2017-10-11 8:44 UTC (permalink / raw)
To: Arnd Bergmann, Greg Kroah-Hartman, Kishon Vijay Abraham I
Cc: linux-pci, linux-omap, linux-kernel, nsekhar
sscanf(misc_device->name, DRV_MODULE_NAME ".%d", &id) in
pci_endpoint_test_remove returns '0' which results in
pci_endpoint_test_remove returning early without releasing the resources.
This is as a result of misc_device not having a valid name. Fix it here.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
---
drivers/misc/pci_endpoint_test.c | 12 ++++++++++--
1 file changed, 10 insertions(+), 2 deletions(-)
diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c
index e787a63a321a..5cb624b6fa0a 100644
--- a/drivers/misc/pci_endpoint_test.c
+++ b/drivers/misc/pci_endpoint_test.c
@@ -551,17 +551,24 @@ static int pci_endpoint_test_probe(struct pci_dev *pdev,
snprintf(name, sizeof(name), DRV_MODULE_NAME ".%d", id);
misc_device = &test->miscdev;
misc_device->minor = MISC_DYNAMIC_MINOR;
- misc_device->name = name;
+ misc_device->name = kstrdup(name, GFP_KERNEL);
+ if (!misc_device->name) {
+ err = -ENOMEM;
+ goto err_ida_remove;
+ }
misc_device->fops = &pci_endpoint_test_fops,
err = misc_register(misc_device);
if (err) {
dev_err(dev, "failed to register device\n");
- goto err_ida_remove;
+ goto err_kfree_name;
}
return 0;
+err_kfree_name:
+ kfree(misc_device->name);
+
err_ida_remove:
ida_simple_remove(&pci_endpoint_test_ida, id);
@@ -592,6 +599,7 @@ static void pci_endpoint_test_remove(struct pci_dev *pdev)
return;
misc_deregister(&test->miscdev);
+ kfree(misc_device->name);
ida_simple_remove(&pci_endpoint_test_ida, id);
for (bar = BAR_0; bar <= BAR_5; bar++) {
if (test->bar[bar])
--
2.11.0
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH 3/3] misc: pci_endpoint_test: Fix BUG_ON error during pci_disable_msi
2017-10-11 8:44 [PATCH 0/3] pci_endpoint_test: Fix module removal/insertion cycle Kishon Vijay Abraham I
2017-10-11 8:44 ` [PATCH 1/3] misc: pci_endpoint_test: Fix failure path return values in probe Kishon Vijay Abraham I
2017-10-11 8:44 ` [PATCH 2/3] misc: pci_endpoint_test: Fix pci_endpoint_test not releasing resources on remove Kishon Vijay Abraham I
@ 2017-10-11 8:44 ` Kishon Vijay Abraham I
2017-10-17 19:16 ` [PATCH 0/3] pci_endpoint_test: Fix module removal/insertion cycle Bjorn Helgaas
3 siblings, 0 replies; 5+ messages in thread
From: Kishon Vijay Abraham I @ 2017-10-11 8:44 UTC (permalink / raw)
To: Arnd Bergmann, Greg Kroah-Hartman, Kishon Vijay Abraham I
Cc: linux-pci, linux-omap, linux-kernel, nsekhar
pci_disable_msi throws a Kernel BUG if the driver has
successfully requested an irq and not released it. Fix it here by freeing
irqs before invoking pci_disable_msi.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
---
drivers/misc/pci_endpoint_test.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/misc/pci_endpoint_test.c b/drivers/misc/pci_endpoint_test.c
index 5cb624b6fa0a..086a7f85c2c0 100644
--- a/drivers/misc/pci_endpoint_test.c
+++ b/drivers/misc/pci_endpoint_test.c
@@ -92,6 +92,7 @@ struct pci_endpoint_test {
void __iomem *bar[6];
struct completion irq_raised;
int last_irq;
+ int num_irqs;
/* mutex to protect the ioctls */
struct mutex mutex;
struct miscdevice miscdev;
@@ -504,6 +505,7 @@ static int pci_endpoint_test_probe(struct pci_dev *pdev,
irq = pci_alloc_irq_vectors(pdev, 1, 32, PCI_IRQ_MSI);
if (irq < 0)
dev_err(dev, "failed to get MSI interrupts\n");
+ test->num_irqs = irq;
}
err = devm_request_irq(dev, pdev->irq, pci_endpoint_test_irqhandler,
@@ -578,6 +580,9 @@ static int pci_endpoint_test_probe(struct pci_dev *pdev,
pci_iounmap(pdev, test->bar[bar]);
}
+ for (i = 0; i < irq; i++)
+ devm_free_irq(dev, pdev->irq + i, test);
+
err_disable_msi:
pci_disable_msi(pdev);
pci_release_regions(pdev);
@@ -591,6 +596,7 @@ static int pci_endpoint_test_probe(struct pci_dev *pdev,
static void pci_endpoint_test_remove(struct pci_dev *pdev)
{
int id;
+ int i;
enum pci_barno bar;
struct pci_endpoint_test *test = pci_get_drvdata(pdev);
struct miscdevice *misc_device = &test->miscdev;
@@ -605,6 +611,8 @@ static void pci_endpoint_test_remove(struct pci_dev *pdev)
if (test->bar[bar])
pci_iounmap(pdev, test->bar[bar]);
}
+ for (i = 0; i < test->num_irqs; i++)
+ devm_free_irq(&pdev->dev, pdev->irq + i, test);
pci_disable_msi(pdev);
pci_release_regions(pdev);
pci_disable_device(pdev);
--
2.11.0
^ permalink raw reply related [flat|nested] 5+ messages in thread* Re: [PATCH 0/3] pci_endpoint_test: Fix module removal/insertion cycle
2017-10-11 8:44 [PATCH 0/3] pci_endpoint_test: Fix module removal/insertion cycle Kishon Vijay Abraham I
` (2 preceding siblings ...)
2017-10-11 8:44 ` [PATCH 3/3] misc: pci_endpoint_test: Fix BUG_ON error during pci_disable_msi Kishon Vijay Abraham I
@ 2017-10-17 19:16 ` Bjorn Helgaas
3 siblings, 0 replies; 5+ messages in thread
From: Bjorn Helgaas @ 2017-10-17 19:16 UTC (permalink / raw)
To: Kishon Vijay Abraham I
Cc: Arnd Bergmann, Greg Kroah-Hartman, linux-pci, linux-omap,
linux-kernel, nsekhar
On Wed, Oct 11, 2017 at 02:14:35PM +0530, Kishon Vijay Abraham I wrote:
> This series fixes module removal/insertion cycle of pci_endpoint_test.
>
> Without this series, when trying to modprobe pci_endpoint_test after
> rmmod pci_endpoint_test results in the following errors.
>
> pci-endpoint-test 0000:01:00.0: BAR 0: can't reserve [mem 0x20204600-0x202046ff]
> pci-endpoint-test 0000:01:00.0: Cannot obtain PCI resources
> pci-endpoint-test: probe of 0000:01:00.0 failed with error -16
>
> Kishon Vijay Abraham I (3):
> misc: pci_endpoint_test: Fix failure path return values in probe
> misc: pci_endpoint_test: Fix pci_endpoint_test not releasing resources
> on remove
> misc: pci_endpoint_test: Fix BUG_ON error during pci_disable_msi
>
> drivers/misc/pci_endpoint_test.c | 22 ++++++++++++++++++++--
> 1 file changed, 20 insertions(+), 2 deletions(-)
Applied to pci/endpoint for v4.15, thanks!
^ permalink raw reply [flat|nested] 5+ messages in thread