Message ID | 1394447948-14778-3-git-send-email-vinod.koul@intel.com (mailing list archive) |
---|---|
State | Accepted |
Commit | 9068b032d0ff |
Delegated to: | Vinod Koul |
Headers | show |
diff --git a/drivers/dma/pch_dma.c b/drivers/dma/pch_dma.c index 61fdc54..05fa548 100644 --- a/drivers/dma/pch_dma.c +++ b/drivers/dma/pch_dma.c @@ -964,16 +964,16 @@ static void pch_dma_remove(struct pci_dev *pdev) if (pd) { dma_async_device_unregister(&pd->dma); + free_irq(pdev->irq, pd); + list_for_each_entry_safe(chan, _c, &pd->dma.channels, device_node) { pd_chan = to_pd_chan(chan); - tasklet_disable(&pd_chan->tasklet); tasklet_kill(&pd_chan->tasklet); } pci_pool_destroy(pd->pool); - free_irq(pdev->irq, pd); pci_iounmap(pdev, pd->membase); pci_release_regions(pdev); pci_disable_device(pdev);
As discussed in [1] the tasklet_disable is not a proper function for teardown. The driver also uses correct method of tasklet_kill. So remove tasklet_disable In case of this driver we also have to ensure that we disable interrupts and also disable scheduling of tasklet so move up the free_irq() call before tasklet_kill() [1]: http://lwn.net/Articles/588457/ Reported-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Vinod Koul <vinod.koul@intel.com> --- drivers/dma/pch_dma.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-)