2005-02-17 12:47:55

by Davide Rossetti

[permalink] [raw]
Subject: rmmod while module is in use

maybe RTFM...
a module:
- char device driver for..
- a PCI device

any clue as to how to protect from module unloading while there is still
some process opening it??? have I to sleep in the remove_one() pci
driver function till last process closes its file descriptor???

static void __devexit apedev_remove_one(struct pci_dev *pdev)
{
ApeDev* apedev = pci_get_drvdata(pdev);

if(test_bit(APEDEV_FLAG_OPEN, &apedev->flags)) {
PERROR("still open flag on!!! (flags=0x%08x)\n", apedev->flags);

// sleep here till it gets closed...

}
...
}

static struct pci_driver apedev_driver = {
.name = DEVNAME,
.id_table = apedev_pci_tbl,
.probe = apedev_init_one,
.remove = __devexit_p(apedev_remove_one),
};


2005-02-17 13:05:14

by linux-os (Dick Johnson)

[permalink] [raw]
Subject: Re: rmmod while module is in use

On Thu, 17 Feb 2005, Davide Rossetti wrote:

> maybe RTFM...
> a module:
> - char device driver for..
> - a PCI device
>
> any clue as to how to protect from module unloading while there is still some
> process opening it??? have I to sleep in the remove_one() pci driver function
> till last process closes its file descriptor???
>

The kernel code is supposed to prevent module removal when it
is still in use. Have you discovered a bug where the kernel
will allow unloading when it's still being used???

There used to be MOD_INC_USE_COUNT and MOD_DEC_USE_COUNT
macros to be using in open() and close(). However their
use has been depreciated in favor of some internal kernel
magic. So, unless you have discovered a bug, your code
doesn't have to worry anymore.

> static void __devexit apedev_remove_one(struct pci_dev *pdev)
> {
> ApeDev* apedev = pci_get_drvdata(pdev);
>
> if(test_bit(APEDEV_FLAG_OPEN, &apedev->flags)) {
> PERROR("still open flag on!!! (flags=0x%08x)\n", apedev->flags);
>
> // sleep here till it gets closed...
>
> }
> ...
> }
>
> static struct pci_driver apedev_driver = {
> .name = DEVNAME,
> .id_table = apedev_pci_tbl,
> .probe = apedev_init_one,
> .remove = __devexit_p(apedev_remove_one),
> };
>

Cheers,
Dick Johnson
Penguin : Linux version 2.6.10 on an i686 machine (5537.79 BogoMips).
Notice : All mail here is now cached for review by Dictator Bush.
98.36% of all statistics are fiction.

2005-02-17 13:11:06

by Sean Neakums

[permalink] [raw]
Subject: Re: rmmod while module is in use

Davide Rossetti <[email protected]> writes:

> maybe RTFM...
> a module:
> - char device driver for..
> - a PCI device

Setting the 'owner' field of your char device's file_operations
structure to THIS_MODULE should be sufficient to enable the kernel to
manage the reference count for you. This is the "magic" referred to
in linux-os's reply.

2005-02-17 14:39:59

by Zwane Mwaikambo

[permalink] [raw]
Subject: Re: rmmod while module is in use

On Thu, 17 Feb 2005, Davide Rossetti wrote:

> maybe RTFM...
> a module:
> - char device driver for..
> - a PCI device
>
> any clue as to how to protect from module unloading while there is still some
> process opening it??? have I to sleep in the remove_one() pci driver function
> till last process closes its file descriptor???
>
> static void __devexit apedev_remove_one(struct pci_dev *pdev)
> {
> ApeDev* apedev = pci_get_drvdata(pdev);
>
> if(test_bit(APEDEV_FLAG_OPEN, &apedev->flags)) {
> PERROR("still open flag on!!! (flags=0x%08x)\n", apedev->flags);
>
> // sleep here till it gets closed...
>
> }
> ...
> }
>
> static struct pci_driver apedev_driver = {
> .name = DEVNAME,
> .id_table = apedev_pci_tbl,
> .probe = apedev_init_one,
> .remove = __devexit_p(apedev_remove_one),
> };

Add .module = THIS_MODULE to your file_operations.

2005-02-17 14:49:23

by Zwane Mwaikambo

[permalink] [raw]
Subject: Re: rmmod while module is in use

On Thu, 17 Feb 2005, Zwane Mwaikambo wrote:

> On Thu, 17 Feb 2005, Davide Rossetti wrote:
>
> > maybe RTFM...
> > a module:
> > - char device driver for..
> > - a PCI device
> >
> > any clue as to how to protect from module unloading while there is still some
> > process opening it??? have I to sleep in the remove_one() pci driver function
> > till last process closes its file descriptor???
> >
> > static void __devexit apedev_remove_one(struct pci_dev *pdev)
> > {
> > ApeDev* apedev = pci_get_drvdata(pdev);
> >
> > if(test_bit(APEDEV_FLAG_OPEN, &apedev->flags)) {
> > PERROR("still open flag on!!! (flags=0x%08x)\n", apedev->flags);
> >
> > // sleep here till it gets closed...
> >
> > }
> > ...
> > }
> >
> > static struct pci_driver apedev_driver = {
> > .name = DEVNAME,
> > .id_table = apedev_pci_tbl,
> > .probe = apedev_init_one,
> > .remove = __devexit_p(apedev_remove_one),
> > };
>
> Add .module = THIS_MODULE to your file_operations.
^^^^^^^

That should be .owner