Fix double mem region release

Two one-liners for the same issue, second bug occurrence is just a copy&pasted mistake from 2011...

Ladislav Michl (2):
devres: Fix double mem region release in devm_ioremap_resource()
PCI: Fix double mem region release in devm_pci_remap_cfg_resource()

 drivers/pci/pci.c | 1 -
 lib/devres.c      | 1 -
 2 files changed, 2 deletions(-)

2.15.1

devm_release_mem_region() is called explicitely in case devm_ioremap() fails, however the same release function is later called also as devres release of devm_request_mem_region() causing double resource free.

---
 lib/devres.c | 1 -
 1 file changed, 1 deletion(-)

index 5f2aedd58bc5..584356a568d0 100644
@@ -156,7 +156,6 @@ void __iomem *devm_ioremap_resource(struct device *dev, struct resource *res)
 	dest_ptr = devm_ioremap(dev, res->start, size);
 	if (!dest_ptr) {
 		dev_err(dev, "ioremap failed for resource %pR\n", res);
-		devm_release_mem_region(dev, res->start, size);
 		dest_ptr = IOMEM_ERR_PTR(-ENOMEM);
 	}
 

2.15.1

devm_release_mem_region() is called explicitely in case devm_pci_remap_cfgspace() fails, however the same release function is later called also as devres release of devm_request_mem_region() causing double resource free.

---
 drivers/pci/pci.c | 1 -
 1 file changed, 1 deletion(-)

index 764ca7b8840d..8f9d81a23ca5 100644
@@ -3607,7 +3607,6 @@ void __iomem *devm_pci_remap_cfg_resource(struct device *dev,
 	dest_ptr = devm_pci_remap_cfgspace(dev, res->start, size);
 	if (!dest_ptr) {
 		dev_err(dev, "ioremap failed for resource %pR\n", res);
-		devm_release_mem_region(dev, res->start, size);
 		dest_ptr = IOMEM_ERR_PTR(-ENOMEM);
 	}
 

2.15.1

Hmm, that is really shame :-/ My test setup triggered "Trying to free nonexistent resource" warning, but dissasembly of lib/devres.o shows no call to devres_destroy, which seems that WARN_ON in __devm_release_region() is optimized out. Strange. So far sorry for the noise, I'll try figure out what broke after weekend.