Patchwork drivers/mtd/nand/pasemi_nand.c: Add missing pci_dev_put

login
register
mail settings
Submitter Julia Lawall
Date Nov. 27, 2008, 12:19 p.m.
Message ID <Pine.LNX.4.64.0811271319070.15160@pc-004.diku.dk>
Download mbox | patch
Permalink /patch/11176/
State New
Headers show

Comments

Julia Lawall - Nov. 27, 2008, 12:19 p.m.
From: Julia Lawall <julia@diku.dk>

pci_get_device increments a reference count that should be decremented
using pci_dev_put.  I have thus added an extra label in the error handling
code to do this.  I don't know, however, whether there should be a
pci_dev_put before the return 0 as well.

The semantic match that finds the problem is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@r exists@
local idexpression x;
statement S,S1;
position p1,p2,p3;
expression E,E1;
type T,T1;
expression *ptr != NULL;
@@

(
 if ((x@p1 = pci_get_device(...)) == NULL) S
|
 x@p1 = pci_get_device(...);
)
 ... when != pci_dev_put(...,(T)x,...)
     when != if (...) { <+... pci_dev_put(...,(T)x,...) ...+> }
     when != true x == NULL || ...
     when != x = E
     when != E = (T)x
     when any
(
 if (x == NULL || ...) S1
|
 if@p2 (...) {
  ... when != pci_dev_put(...,(T1)x,...)
      when != if (...) { <+... pci_dev_put(...,(T1)x,...) ...+> }
      when != x = E1
      when != E1 = (T1)x
(
  return \(0\|<+...x...+>\|ptr\);
|
  return@p3 ...;
)
}
)

@ script:python @
p1 << r.p1;
p3 << r.p3;
@@

print "* file: %s pci_get_device: %s return: %s" % (p1[0].file,p1[0].line,p3[0].line)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>

---
 drivers/mtd/nand/pasemi_nand.c      |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)
Julia Lawall - Dec. 1, 2008, 5:39 p.m.
On Mon, 1 Dec 2008, Olof Johansson wrote:

> On Thu, Nov 27, 2008 at 01:19:49PM +0100, Julia Lawall wrote:
> > From: Julia Lawall <julia@diku.dk>
> > 
> > pci_get_device increments a reference count that should be decremented
> > using pci_dev_put.  I have thus added an extra label in the error handling
> > code to do this.  I don't know, however, whether there should be a
> > pci_dev_put before the return 0 as well.
> > 
> [...]
> 
> > Signed-off-by: Julia Lawall <julia@diku.dk>
> 
> Acked-by: Olof Johansson <olof@lixom.net>

Do you know the answer about the return 0?

julia
Olof Johansson - Dec. 1, 2008, 5:43 p.m.
On Thu, Nov 27, 2008 at 01:19:49PM +0100, Julia Lawall wrote:
> From: Julia Lawall <julia@diku.dk>
> 
> pci_get_device increments a reference count that should be decremented
> using pci_dev_put.  I have thus added an extra label in the error handling
> code to do this.  I don't know, however, whether there should be a
> pci_dev_put before the return 0 as well.
> 
[...]

> Signed-off-by: Julia Lawall <julia@diku.dk>

Acked-by: Olof Johansson <olof@lixom.net>
Julia Lawall - Dec. 1, 2008, 6:03 p.m.
On Mon, 1 Dec 2008, Olof Johansson wrote:

> On Mon, Dec 01, 2008 at 06:39:01PM +0100, Julia Lawall wrote:
> > On Mon, 1 Dec 2008, Olof Johansson wrote:
> > 
> > > On Thu, Nov 27, 2008 at 01:19:49PM +0100, Julia Lawall wrote:
> > > > From: Julia Lawall <julia@diku.dk>
> > > > 
> > > > pci_get_device increments a reference count that should be decremented
> > > > using pci_dev_put.  I have thus added an extra label in the error handling
> > > > code to do this.  I don't know, however, whether there should be a
> > > > pci_dev_put before the return 0 as well.
> > > > 
> > > [...]
> > > 
> > > > Signed-off-by: Julia Lawall <julia@diku.dk>
> > > 
> > > Acked-by: Olof Johansson <olof@lixom.net>
> > 
> > Do you know the answer about the return 0?
> 
> Teaches me to read the patch description twice.
> 
> Either there or in pasemi_nand_remove(), doesn't matter much to me. Doing
> it before the return 0 is the smaller change.

OK, I will send an adjusted patch shortly.

Thanks,
julia
Olof Johansson - Dec. 1, 2008, 6:08 p.m.
On Mon, Dec 01, 2008 at 06:39:01PM +0100, Julia Lawall wrote:
> On Mon, 1 Dec 2008, Olof Johansson wrote:
> 
> > On Thu, Nov 27, 2008 at 01:19:49PM +0100, Julia Lawall wrote:
> > > From: Julia Lawall <julia@diku.dk>
> > > 
> > > pci_get_device increments a reference count that should be decremented
> > > using pci_dev_put.  I have thus added an extra label in the error handling
> > > code to do this.  I don't know, however, whether there should be a
> > > pci_dev_put before the return 0 as well.
> > > 
> > [...]
> > 
> > > Signed-off-by: Julia Lawall <julia@diku.dk>
> > 
> > Acked-by: Olof Johansson <olof@lixom.net>
> 
> Do you know the answer about the return 0?

Teaches me to read the patch description twice.

Either there or in pasemi_nand_remove(), doesn't matter much to me. Doing
it before the return 0 is the smaller change.


-Olof
Julia Lawall - Dec. 1, 2008, 8:11 p.m.
On Mon, 1 Dec 2008, Olof Johansson wrote:

> On Mon, Dec 01, 2008 at 06:39:01PM +0100, Julia Lawall wrote:
> > On Mon, 1 Dec 2008, Olof Johansson wrote:
> > 
> > > On Thu, Nov 27, 2008 at 01:19:49PM +0100, Julia Lawall wrote:
> > > > From: Julia Lawall <julia@diku.dk>
> > > > 
> > > > pci_get_device increments a reference count that should be decremented
> > > > using pci_dev_put.  I have thus added an extra label in the error handling
> > > > code to do this.  I don't know, however, whether there should be a
> > > > pci_dev_put before the return 0 as well.
> > > > 
> > > [...]
> > > 
> > > > Signed-off-by: Julia Lawall <julia@diku.dk>
> > > 
> > > Acked-by: Olof Johansson <olof@lixom.net>
> > 
> > Do you know the answer about the return 0?
> 
> Teaches me to read the patch description twice.
> 
> Either there or in pasemi_nand_remove(), doesn't matter much to me. Doing
> it before the return 0 is the smaller change.

Actually, could pci_dev_put(pdev) be placed right after:

lpcctl = pci_resource_start(pdev, 0);

Then there would only be one, and the rest of the code could go back to 
its original form.

julia
Olof Johansson - Dec. 1, 2008, 9:57 p.m.
On Mon, Dec 01, 2008 at 09:11:09PM +0100, Julia Lawall wrote:
> On Mon, 1 Dec 2008, Olof Johansson wrote:
> 
> > On Mon, Dec 01, 2008 at 06:39:01PM +0100, Julia Lawall wrote:
> > > On Mon, 1 Dec 2008, Olof Johansson wrote:
> > > 
> > > > On Thu, Nov 27, 2008 at 01:19:49PM +0100, Julia Lawall wrote:
> > > > > From: Julia Lawall <julia@diku.dk>
> > > > > 
> > > > > pci_get_device increments a reference count that should be decremented
> > > > > using pci_dev_put.  I have thus added an extra label in the error handling
> > > > > code to do this.  I don't know, however, whether there should be a
> > > > > pci_dev_put before the return 0 as well.
> > > > > 
> > > > [...]
> > > > 
> > > > > Signed-off-by: Julia Lawall <julia@diku.dk>
> > > > 
> > > > Acked-by: Olof Johansson <olof@lixom.net>
> > > 
> > > Do you know the answer about the return 0?
> > 
> > Teaches me to read the patch description twice.
> > 
> > Either there or in pasemi_nand_remove(), doesn't matter much to me. Doing
> > it before the return 0 is the smaller change.
> 
> Actually, could pci_dev_put(pdev) be placed right after:
> 
> lpcctl = pci_resource_start(pdev, 0);
> 
> Then there would only be one, and the rest of the code could go back to 
> its original form.

Yep, even easier.

Thanks for catching this.


-Olof

Patch

diff --git a/drivers/mtd/nand/pasemi_nand.c b/drivers/mtd/nand/pasemi_nand.c
index 75c8990..ceae196 100644
--- a/drivers/mtd/nand/pasemi_nand.c
+++ b/drivers/mtd/nand/pasemi_nand.c
@@ -144,7 +144,7 @@  static int __devinit pasemi_nand_probe(struct of_device *ofdev,
 
 	if (!request_region(lpcctl, 4, driver_name)) {
 		err = -EBUSY;
-		goto out_ior;
+		goto out_pdev;
 	}
 
 	chip->cmd_ctrl = pasemi_hwcontrol;
@@ -176,6 +176,8 @@  static int __devinit pasemi_nand_probe(struct of_device *ofdev,
 
  out_lpc:
 	release_region(lpcctl, 4);
+ out_pdev:
+	pci_dev_put(pdev);
  out_ior:
 	iounmap(chip->IO_ADDR_R);
  out_mtd: