2008-08-29 21:46:41

by Elias Oltmanns

[permalink] [raw]
Subject: [PATCH] ide: Two fixes regarding memory allocation

In function ide_devset_execute() we should use __GFP_WAIT rather than
GFP_KERNEL. Also, the allocation cannot possibly fail at that point.
More importantly, there is a potential memory leak in the device probing
code. The infrastructure seems rather complex and I hope I haven't messed
anything up by trying to fix this.

Signed-off-by: Elias Oltmanns <[email protected]>
---
Applies to next-20080829.

drivers/ide/ide-io.c | 5 +----
drivers/ide/ide-probe.c | 22 +++++++++++++++++-----
2 files changed, 18 insertions(+), 9 deletions(-)

diff --git a/drivers/ide/ide-io.c b/drivers/ide/ide-io.c
index d0579f1..e205f46 100644
--- a/drivers/ide/ide-io.c
+++ b/drivers/ide/ide-io.c
@@ -655,10 +655,7 @@ int ide_devset_execute(ide_drive_t *drive, const struct ide_devset *setting,
if (!(setting->flags & DS_SYNC))
return setting->set(drive, arg);

- rq = blk_get_request(q, READ, GFP_KERNEL);
- if (!rq)
- return -ENOMEM;
-
+ rq = blk_get_request(q, READ, __GFP_WAIT);
rq->cmd_type = REQ_TYPE_SPECIAL;
rq->cmd_len = 5;
rq->cmd[0] = REQ_DEVSET_EXEC;
diff --git a/drivers/ide/ide-probe.c b/drivers/ide/ide-probe.c
index b5e54d2..1c4cd5e 100644
--- a/drivers/ide/ide-probe.c
+++ b/drivers/ide/ide-probe.c
@@ -958,9 +958,9 @@ static void ide_add_drive_to_hwgroup(ide_drive_t *drive)
* - allocate the block device queue
* - link drive into the hwgroup
*/
-static void ide_port_setup_devices(ide_hwif_t *hwif)
+static int ide_port_setup_devices(ide_hwif_t *hwif)
{
- int i;
+ int i, j = 0;

mutex_lock(&ide_cfg_mtx);
for (i = 0; i < MAX_DRIVES; i++) {
@@ -972,12 +972,21 @@ static void ide_port_setup_devices(ide_hwif_t *hwif)
if (ide_init_queue(drive)) {
printk(KERN_ERR "ide: failed to init %s\n",
drive->name);
+ spin_lock_irq(&ide_lock);
+ kfree(drive->id);
+ drive->id = NULL;
+ drive->dev_flags &= ~IDE_DFLAG_PRESENT;
+ spin_unlock_irq(&ide_lock);
continue;
}

+ j++;
+
ide_add_drive_to_hwgroup(drive);
}
mutex_unlock(&ide_cfg_mtx);
+
+ return j;
}

static ide_hwif_t *ide_ports[MAX_HWIFS];
@@ -1663,10 +1672,13 @@ int ide_host_register(struct ide_host *host, const struct ide_port_info *d,
continue;
}

- j++;
-
if (hwif->present)
- ide_port_setup_devices(hwif);
+ if (ide_port_setup_devices(hwif) == 0) {
+ hwif->present = 0;
+ continue;
+ }
+
+ j++;

ide_acpi_init(hwif);



Subject: Re: [PATCH] ide: Two fixes regarding memory allocation

On Friday 29 August 2008, Elias Oltmanns wrote:
> In function ide_devset_execute() we should use __GFP_WAIT rather than
> GFP_KERNEL. Also, the allocation cannot possibly fail at that point.
> More importantly, there is a potential memory leak in the device probing
> code. The infrastructure seems rather complex and I hope I haven't messed
> anything up by trying to fix this.
>
> Signed-off-by: Elias Oltmanns <[email protected]>

thanks, applied

> @@ -972,12 +972,21 @@ static void ide_port_setup_devices(ide_hwif_t *hwif)
> if (ide_init_queue(drive)) {
> printk(KERN_ERR "ide: failed to init %s\n",
> drive->name);
> + spin_lock_irq(&ide_lock);
> + kfree(drive->id);
> + drive->id = NULL;
> + drive->dev_flags &= ~IDE_DFLAG_PRESENT;
> + spin_unlock_irq(&ide_lock);
> continue;

ide_lock taking is superfluous here, I removed it while merging the patch