On Thu, Apr 03, 2014 at 03:12:54PM -0600, Bjorn Helgaas wrote: >On Thu, Apr 03, 2014 at 02:54:32PM +0800, Wei Yang wrote: >> Here is another one based on your comment, which split mlx4_remove_one into >> two and named a helper __mlx4_remove_one(). mlx4_pci_err_detected() will just >> call __mlx4_remove_one(), which will not release drvdata. >> >> BTW, this is not tested, just want to make sure my understanding is correct. > >A couple minor comments below, but in general, yes, this is what I was >thinking. > >> From 84a5a9df0604cbea9b70c74b0709258841637946 Mon Sep 17 00:00:00 2001 >> From: Wei Yang <weiyang@xxxxxxxxxxxxxxxxxx> >> Date: Mon, 31 Mar 2014 11:34:57 +0800 >> Subject: [PATCH] net/mlx4_core: match pci_device_id including dynids >> >> Fix issue introduced by commit: 97a5221 "net/mlx4_core: pass >> pci_device_id.driver_data to __mlx4_init_one during reset". >> >> pci_match_id() just match the static pci_device_id, which may return NULL if >> someone binds the driver to a device manually using >> /sys/bus/pci/drivers/.../new_id. >> >> This patch wrap up a helper function __mlx4_remove_one() which does the tear >> down function but preserve the drv_data. Functions like >> mlx4_pci_err_detected() and mlx4_restart_one() will call this one with out >> releasing drvdata. >> >> CC: Bjorn Helgaas <bhelgaas@xxxxxxxxxx> >> CC: Amir Vadai <amirv@xxxxxxxxxxxx> >> Signed-off-by: Wei Yang <weiyang@xxxxxxxxxxxxxxxxxx> >> Acked-by: Amir Vadai <amirv@xxxxxxxxxxxx> >> --- >> drivers/net/ethernet/mellanox/mlx4/main.c | 149 ++++++++++++++++------------- >> 1 file changed, 80 insertions(+), 69 deletions(-) >> >> diff --git a/drivers/net/ethernet/mellanox/mlx4/main.c b/drivers/net/ethernet/mellanox/mlx4/main.c >> index aa54ef7..fd1f288 100644 >> --- a/drivers/net/ethernet/mellanox/mlx4/main.c >> +++ b/drivers/net/ethernet/mellanox/mlx4/main.c >> @@ -2268,7 +2268,12 @@ static int __mlx4_init_one(struct pci_dev *pdev, int pci_dev_data) >> /* Allow large DMA segments, up to the firmware limit of 1 GB */ >> dma_set_max_seg_size(&pdev->dev, 1024 * 1024 * 1024); >> >> - priv = kzalloc(sizeof(*priv), GFP_KERNEL); >> + dev = pci_get_drvdata(pdev); >> + if (!dev) >> + priv = kzalloc(sizeof(*priv), GFP_KERNEL); >> + else >> + priv = mlx4_priv(dev); > >Why don't you move the priv kzalloc into mlx4_init_one()? Then it would be >symmetric -- you alloc and call pci_set_drvdata() in mlx4_init_one(), and >you call pci_set_drvdata(NULL) and free it in mlx4_remove_one(). And you >wouldn't need the test here. > Agree, this looks more consistent. Will write a formal version and send to mail list after verification. -- Richard Yang Help you, Help me -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html