On Fri, 2010-04-16 at 15:34 +0800, Li Yang wrote: > From: Zhao Chenhui <b26...@freescale.com> > > Put all fsl_msi banks in a linked list. The list of banks then can be > traversed when allocating new msi interrupts.
So there are multiple banks, and you just use the first one that has an empty slot in it's bitmap? > Signed-off-by: Zhao Chenhui <b26...@freescale.com> > Signed-off-by: Li Yang <le...@freescale.com> > --- > arch/powerpc/sysdev/fsl_msi.c | 29 ++++++++++++++++++++++------- > arch/powerpc/sysdev/fsl_msi.h | 2 ++ > 2 files changed, 24 insertions(+), 7 deletions(-) > > diff --git a/arch/powerpc/sysdev/fsl_msi.c b/arch/powerpc/sysdev/fsl_msi.c > index 716862f..c46db75 100644 > --- a/arch/powerpc/sysdev/fsl_msi.c > +++ b/arch/powerpc/sysdev/fsl_msi.c > @@ -24,6 +24,8 @@ > #include <asm/mpic.h> > #include "fsl_msi.h" > > +LIST_HEAD(msi_head); > + > struct fsl_msi_feature { > u32 fsl_pic_ip; > u32 msiir_offset; > @@ -143,15 +145,26 @@ static int fsl_setup_msi_irqs(struct pci_dev *pdev, int > nvec, int type) > struct msi_msg msg; > struct fsl_msi *msi_data; > > + if (list_empty(&msi_head)) { > + pr_debug("%s: msi init error\n", __func__); > + rc = -EFAULT; > + goto out_free; > + } If there's an error probing then fsl_setup_msi_irqs() should not be installed as ppc_md.setup_msi_irqs(), and this code should never run. ie. You shouldn't need this check. > + > list_for_each_entry(entry, &pdev->msi_list, list) { > - msi_data = get_irq_chip_data(entry->irq); > > - hwirq = msi_bitmap_alloc_hwirqs(&msi_data->bitmap, 1); > - if (hwirq < 0) { > - rc = hwirq; > - pr_debug("%s: fail allocating msi interrupt\n", > - __func__); > - goto out_free; > + list_for_each_entry(msi_data, &msi_head, list) { > + > + hwirq = msi_bitmap_alloc_hwirqs(&msi_data->bitmap, 1); > + if (hwirq >= 0) > + break; > + > + if (list_is_last(&msi_data->list, &msi_head)) { > + rc = hwirq; > + pr_debug("%s: fail allocating msi interrupt\n", > + __func__); > + goto out_free; > + } You could make this cleaner by pulling the inner loop into a separate function, and when you fall off the end of the list you return < 0. That would avoid needing the list_is_last() check. cheers
signature.asc
Description: This is a digitally signed message part
_______________________________________________ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev