Hi, looking at region and resource handling I see that request_region
allocates the struct resource with kmalloc() while request_resource
takes it as a parameter. This means that release_region uses kfree()
and release_resource doesn't.
There are some drivers which mix this up by doing a request_region /
release_resource pair, instead of the proper request_region /
release_region. This leads to a memory leak.
I've tried to fix this below. Comments? Output is from bk diff -u,
not sure what the best command would've been.
Also, one check_region was removed, and isapnp seems to release the
wrong region, isapnp_rdp instead of _PNPWRP.
Marcus
===== drivers/pnp/isapnp.c 1.7 vs 1.8 =====
--- 1.7/drivers/pnp/isapnp.c Tue Feb 5 16:10:27 2002
+++ 1.8/drivers/pnp/isapnp.c Sat Feb 9 03:40:41 2002
@@ -2148,12 +2148,12 @@
{
#ifdef ISAPNP_REGION_OK
if (pidxr_res)
- release_resource(pidxr_res);
+ release_region(_PIDXR, 1);
#endif
if (pnpwrp_res)
- release_resource(pnpwrp_res);
+ release_region(_PNPWRP, 1);
if (isapnp_rdp >= 0x203 && isapnp_rdp <= 0x3ff && isapnp_rdp_res)
- release_resource(isapnp_rdp_res);
+ release_region(isapnp_rdp, 1);
#ifdef MODULE
#ifdef CONFIG_PROC_FS
isapnp_proc_done();
@@ -2313,7 +2313,7 @@
#ifdef ISAPNP_REGION_OK
release_region(_PIDXR, 1);
#endif
- release_region(isapnp_rdp, 1);
+ release_region(_PNPWRP, 1);
return -EBUSY;
}
isapnp_set_rdp();
===== drivers/net/wd.c 1.6 vs 1.7 =====
--- 1.6/drivers/net/wd.c Tue Feb 5 09:49:26 2002
+++ 1.7/drivers/net/wd.c Sat Feb 9 03:40:41 2002
@@ -97,7 +97,7 @@
return -EBUSY;
i = wd_probe1(dev, base_addr);
if (i != 0)
- release_resource(r);
+ release_region(base_addr, WD_IO_EXTENT);
else
r->name = dev->name;
return i;
@@ -114,7 +114,7 @@
r->name = dev->name;
return 0;
}
- release_resource(r);
+ release_region(ioaddr, WD_IO_EXTENT);
}
return -ENODEV;
===== drivers/net/lance.c 1.8 vs 1.9 =====
--- 1.8/drivers/net/lance.c Tue Feb 5 09:49:26 2002
+++ 1.9/drivers/net/lance.c Sat Feb 9 03:40:41 2002
@@ -383,7 +383,7 @@
return 0;
}
}
- release_resource(r);
+ release_region(ioaddr, LANCE_TOTAL_SIZE);
}
}
return -ENODEV;
===== drivers/net/pcnet32.c 1.14 vs 1.15 =====
--- 1.14/drivers/net/pcnet32.c Tue Feb 5 16:10:21 2002
+++ 1.15/drivers/net/pcnet32.c Sat Feb 9 03:40:41 2002
@@ -704,7 +704,7 @@
/* pci_alloc_consistent returns page-aligned memory, so we do not have to check the alignment */
if ((lp = pci_alloc_consistent(pdev, sizeof(*lp), &lp_dma_addr)) == NULL) {
- release_resource(res);
+ release_region(ioaddr, PCNET32_TOTAL_SIZE);
return -ENOMEM;
}
===== drivers/char/tpqic02.c 1.6 vs 1.7 =====
--- 1.6/drivers/char/tpqic02.c Tue Feb 5 09:45:20 2002
+++ 1.7/drivers/char/tpqic02.c Sat Feb 9 03:40:41 2002
@@ -2750,7 +2750,9 @@
* the config parameters have been set using MTSETCONFIG.
*/
- if (check_region(QIC02_TAPE_PORT, QIC02_TAPE_PORT_RANGE)) {
+ /* Grab the IO region. */
+ if (!request_region(QIC02_TAPE_PORT, QIC02_TAPE_PORT_RANGE,
+ TPQIC02_NAME)) {
printk(TPQIC02_NAME
": IO space at 0x%x [%d ports] already reserved\n",
QIC02_TAPE_PORT, QIC02_TAPE_PORT_RANGE);
@@ -2764,6 +2766,7 @@
printk(TPQIC02_NAME
": can't allocate IRQ%d for QIC-02 tape\n",
QIC02_TAPE_IRQ);
+ release_region(QIC02_TAPE_PORT, QIC02_TAPE_PORT_RANGE);
return -EBUSY;
}
@@ -2773,12 +2776,9 @@
": can't allocate DMA%d for QIC-02 tape\n",
QIC02_TAPE_DMA);
free_irq(QIC02_TAPE_IRQ, NULL);
+ release_region(QIC02_TAPE_PORT, QIC02_TAPE_PORT_RANGE);
return -EBUSY;
}
-
- /* Grab the IO region. We already made sure it's available. */
- request_region(QIC02_TAPE_PORT, QIC02_TAPE_PORT_RANGE,
- TPQIC02_NAME);
/* Setup the page-address for the dma transfer. */
buffaddr =
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Feb 15 2002 - 21:00:27 EST