Re: [PATCH, v2] hwmon: coretemp: use list instead of fixed size array for temp data

["Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>]

[-- Attachment #1: Type: text/plain, Size: 3381 bytes --]

On Fri, May 04, 2012 at 06:34:19AM -0700, Guenter Roeck wrote:
> On Fri, May 04, 2012 at 02:46:06AM -0400, Kirill A. Shutemov wrote:
> > On Thu, May 03, 2012 at 10:41:22PM -0700, Guenter Roeck wrote:
> > > On Thu, May 03, 2012 at 07:18:56AM -0400, Kirill A. Shutemov wrote:
> > > > From: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
> > > > 
> > > > Let's rework code to allow arbitrary number of cores on a CPU, not
> > > > limited by hardcoded array size.
> > > > 
> > > > Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> > > > ---
> > > >  v2:
> > > >    - fix NULL pointer dereference. Thanks to R, Durgadoss;
> > > >    - use mutex instead of spinlock for list locking.
> > > > ---
> > > >  drivers/hwmon/coretemp.c |  178 +++++++++++++++++++++++++++++++++-------------
> > > >  1 files changed, 129 insertions(+), 49 deletions(-)
> > > > 
> > > > diff --git a/drivers/hwmon/coretemp.c b/drivers/hwmon/coretemp.c
> > > > index 54a70fe..1c66131 100644
> > > > --- a/drivers/hwmon/coretemp.c
> > > > +++ b/drivers/hwmon/coretemp.c
> > > > @@ -36,6 +36,8 @@
> > > >  #include <linux/cpu.h>
> > > >  #include <linux/pci.h>
> > > >  #include <linux/smp.h>
> > > > +#include <linux/list.h>
> > > > +#include <linux/kref.h>
> > > >  #include <linux/moduleparam.h>
> > > >  #include <asm/msr.h>
> > > >  #include <asm/processor.h>
> > > > @@ -52,11 +54,9 @@ module_param_named(tjmax, force_tjmax, int, 0444);
> > > >  MODULE_PARM_DESC(tjmax, "TjMax value in degrees Celsius");
> > > > 
> > > >  #define BASE_SYSFS_ATTR_NO     2       /* Sysfs Base attr no for coretemp */
> > > > -#define NUM_REAL_CORES         16      /* Number of Real cores per cpu */
> > > >  #define CORETEMP_NAME_LENGTH   17      /* String Length of attrs */
> > > >  #define MAX_CORE_ATTRS         4       /* Maximum no of basic attrs */
> > > >  #define TOTAL_ATTRS            (MAX_CORE_ATTRS + 1)
> > > > -#define MAX_CORE_DATA          (NUM_REAL_CORES + BASE_SYSFS_ATTR_NO)
> > > > 
> > > >  #define TO_PHYS_ID(cpu)                (cpu_data(cpu).phys_proc_id)
> > > >  #define TO_CORE_ID(cpu)                (cpu_data(cpu).cpu_core_id)
> > > > @@ -82,6 +82,9 @@ MODULE_PARM_DESC(tjmax, "TjMax value in degrees Celsius");
> > > >   * @valid: If this is 1, the current temperature is valid.
> > > >   */
> > > >  struct temp_data {
> > > > +       struct list_head list;
> > > > +       struct kref refcount;
> > > 
> > > Hi,
> > > 
> > > the kref is not needed. The attribute access functions don't
> > > need to be protected since the attributes for a core are deleted
> > > before the core data itself is deleted. So it is not neccessary 
> > > to hold a lock while accessing/using temp_data in the attribute
> > > access functions. All you need is to hold a mutex while you are
> > > manipulating or walking the list.
> > 
> > Without kref, what prevents following situation:
> > 
> > 		CPU-A				CPU-B
> > 	tdata = get_temp_data();
> > 					coretemp_remove_core() {
> > 					    device_remove_file();
> > 					    kfree(tdata);
> > 					}
> > 	<tdata dereference>
> > 
> The remove function requires a semaphore which is held by the access function,
> so device_remove_file() will only proceed after CPU-A is done with the sysfs access.

Understood. I'll update the patch.

-- 
 Kirill A. Shutemov

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]