libata: clear drvdata in ata_host_release()
PATCH 1/1 libata: clear drvdata in ata_host_release()
ata_host_release() uses drvdata to determine ata_host to release and clearing drvdata in ->remove_one causes NULL pointer deference. Clear drvdata only in ata_host_release() after all resources are freed.
This bug was first analyzed by Alan Cox for pata_pcmcia.
---
Alan, this is the bug you talked about, right? I took the fix from "[BUG] PATA_PCMCIA does not work" and expanded it. If this isn't the one, please let me know. Thanks.
drivers/ata/libata-core.c +2 -1
@@ -5638,6 +5638,8 @@ static void ata_host_release(struct device *gendev, void *res)if (host->ops->host_stop) host->ops->host_stop (host); + +dev_set_drvdata (gendev,NULL ); }/**
@@ -5860,7 +5862,6 @@ int ata_device_add(const struct ata_probe_ent *ent) err_out:devres_release_group (dev, ata_device_add); -dev_set_drvdata (dev,NULL );VPRINTK ("EXIT, returning %d\n" , rc);return 0 ; }
drivers/ata/pata_cs5520.c +0 -1
@@ -306,7 +306,6 @@ static void __devexit cs5520_remove_one(struct pci_dev *pdev)struct ata_host *host =dev_get_drvdata (dev);ata_host_detach (host); -dev_set_drvdata (dev,NULL ); }/**
drivers/ata/pata_isapnp.c +0 -1
@@ -128,7 +128,6 @@ static void isapnp_remove_one(struct pnp_dev *idev)struct ata_host *host =dev_get_drvdata (dev);ata_host_detach (host); -dev_set_drvdata (dev,NULL ); }static struct pnp_device_id isapnp_devices[] = {
drivers/ata/pata_pcmcia.c +0 -1
@@ -308,7 +308,6 @@ static void pcmcia_remove_one(struct pcmcia_device *pdev)if (info->ndev) {struct ata_host *host =dev_get_drvdata (dev);ata_host_detach (host); -dev_set_drvdata (dev,NULL ); } info->ndev =0 ; pdev->priv =NULL ;
drivers/ata/pata_platform.c +0 -1
@@ -228,7 +228,6 @@ static int __devexit pata_platform_remove(struct platform_device *pdev)struct ata_host *host =dev_get_drvdata (dev);ata_host_detach (host); -dev_set_drvdata (dev,NULL );return 0 ; }
Quoted 9 lines · On Sat, 24 Feb 2007 16:48:23 +0900 Tejun Heo <[email protected]> wrote:
ata_host_release()uses drvdata to determineata_hostto release and clearing drvdata in ->remove_one causes NULL pointer deference. Clear drvdata only inata_host_release()after all resources are freed.This bug was first analyzed by Alan Cox for
pata_pcmcia.Signed-off-by Tejun Heo <[email protected]>Cc Alan Cox <[email protected]>
Quoted 4 lines · Tejun Heo
ata_host_release()uses drvdata to determineata_hostto release and clearing drvdata in ->remove_one causes NULL pointer deference. Clear drvdata only inata_host_release()after all resources are freed.
Quoted 8 lines
This bug was first analyzed by Alan Cox for
pata_pcmcia.Signed-off-by Tejun Heo <[email protected]>Cc Alan Cox <[email protected]>---
Alan, this is the bug you talked about, right? I took the fix from "[BUG]PATA_PCMCIAdoes not work" and expanded it. If this isn't the one, please let me know. Thanks.
Has a tester confirmed that this patch fixes the problem, and supercedes Alan's patch?
Jeff
Quoted 14 lines · On Sat, 24 Feb 2007 20:51:27 -0500 Jeff Garzik <[email protected]> wrote:
Quoted 4 lines · Tejun Heo
ata_host_release()uses drvdata to determineata_hostto release and clearing drvdata in ->remove_one causes NULL pointer deference. Clear drvdata only inata_host_release()after all resources are freed.Quoted 8 lines
This bug was first analyzed by Alan Cox for
pata_pcmcia.Signed-off-by Tejun Heo <[email protected]>Cc Alan Cox <[email protected]>---
Alan, this is the bug you talked about, right? I took the fix from "[BUG]PATA_PCMCIAdoes not work" and expanded it. If this isn't the one, please let me know. Thanks.
Quoted 2 lines
Has a tester confirmed that this patch fixes the problem, and supercedes Alan's patch?
It doesn't appear to, instead it fixes further cases.
PATCH 1/1 libata: clear drvdata in ata_host_release(), take#2
In reply to Jeff GarzikClearing drvdata in ->remove_one causes NULL pointer deference. Clear drvdata only in ata_host_release() after all resources are freed.
---pata_pcmcia is fixed by Alan's patch. This fixes other drivers and make libata always clear drvdata after detach.
drivers/ata/libata-core.c +2 -1
@@ -5680,6 +5680,8 @@ static void ata_host_release(struct device *gendev, void *res)if (host->ops->host_stop) host->ops->host_stop (host); + +dev_set_drvdata (gendev,NULL ); }/**
@@ -5902,7 +5904,6 @@ int ata_device_add(const struct ata_probe_ent *ent) err_out:devres_release_group (dev, ata_device_add); -dev_set_drvdata (dev,NULL );VPRINTK ("EXIT, returning %d\n" , rc);return 0 ; }
drivers/ata/pata_cs5520.c +0 -1
@@ -306,7 +306,6 @@ static void __devexit cs5520_remove_one(struct pci_dev *pdev)struct ata_host *host =dev_get_drvdata (dev);ata_host_detach (host); -dev_set_drvdata (dev,NULL ); }/**
drivers/ata/pata_isapnp.c +0 -1
@@ -128,7 +128,6 @@ static void isapnp_remove_one(struct pnp_dev *idev)struct ata_host *host =dev_get_drvdata (dev);ata_host_detach (host); -dev_set_drvdata (dev,NULL ); }static struct pnp_device_id isapnp_devices[] = {
drivers/ata/pata_platform.c +0 -1
@@ -228,7 +228,6 @@ static int __devexit pata_platform_remove(struct platform_device *pdev)struct ata_host *host =dev_get_drvdata (dev);ata_host_detach (host); -dev_set_drvdata (dev,NULL );return 0 ; }
Jeff Garzik
AppliedRe: [PATCH] libata: clear drvdata in ata_host_release(), take#2
Quoted 7 lines · Tejun Heo
Clearing drvdata in ->remove_one causes NULL pointer deference. Clear drvdata only in
ata_host_release()after all resources are freed.Signed-off-by Tejun Heo <[email protected]>---
pata_pcmciais fixed by Alan's patch. This fixes other drivers and make libata always clear drvdata after detach.
applied