On Thu, Apr 23, 2020 at 2:28 AM Gustavo Pimentel <Gustavo.Pimentel@xxxxxxxxxxxx> wrote: > > On Thu, Apr 23, 2020 at 2:58:21, Alan Mikhak <alan.mikhak@xxxxxxxxxx> > wrote: > > > From: Alan Mikhak <alan.mikhak@xxxxxxxxxx> > > > > Modify dw_edma_irq_request() to check if a struct msi_desc entry exists > > before copying the contents of its struct msi_msg pointer. > > > > Without this sanity check, __get_cached_msi_msg() crashes when invoked by > > dw_edma_irq_request() running on a Linux-based PCIe endpoint device. MSI > > interrupt are not received by PCIe endpoint devices. If irq_get_msi_desc() > > returns null, then there is no cached struct msi_msg to be copied. > > > > This patch depends on the following patch: > > [PATCH v2] dmaengine: dw-edma: Decouple dw-edma-core.c from struct pci_dev > > https://urldefense.com/v3/__https://patchwork.kernel.org/patch/11491757/__;!!A4F2R9G_pg!L_vf_Tml7Ca4sWVvZp5crRCp7YsMj6B93G9cMAO8Dj3w9I0MArjwuwNKtDz9rr0RlpXiqPg$ > > > > Rebased on linux-next which has above patch applied. > > > > Fixes: Build error with config x86_64-randconfig-f003-20200422 > > Fixes: Build error with config s390-allmodconfig > > Reported-by: kbuild test robot <lkp@xxxxxxxxx> > > Signed-off-by: Alan Mikhak <alan.mikhak@xxxxxxxxxx> > > --- > > drivers/dma/dw-edma/dw-edma-core.c | 17 ++++++++++------- > > 1 file changed, 10 insertions(+), 7 deletions(-) > > > > diff --git a/drivers/dma/dw-edma/dw-edma-core.c b/drivers/dma/dw-edma/dw-edma-core.c > > index db401eb11322..306ab50462be 100644 > > --- a/drivers/dma/dw-edma/dw-edma-core.c > > +++ b/drivers/dma/dw-edma/dw-edma-core.c > > @@ -13,6 +13,7 @@ > > #include <linux/dmaengine.h> > > #include <linux/err.h> > > #include <linux/interrupt.h> > > +#include <linux/irq.h> > > #include <linux/dma/edma.h> > > #include <linux/dma-mapping.h> > > > > @@ -773,6 +774,7 @@ static int dw_edma_irq_request(struct dw_edma_chip *chip, > > u32 rd_mask = 1; > > int i, err = 0; > > u32 ch_cnt; > > + int irq; > > > > ch_cnt = dw->wr_ch_cnt + dw->rd_ch_cnt; > > > > @@ -781,16 +783,16 @@ static int dw_edma_irq_request(struct dw_edma_chip *chip, > > > > if (dw->nr_irqs == 1) { > > /* Common IRQ shared among all channels */ > > - err = request_irq(dw->ops->irq_vector(dev, 0), > > - dw_edma_interrupt_common, > > + irq = dw->ops->irq_vector(dev, 0); > > + err = request_irq(irq, dw_edma_interrupt_common, > > IRQF_SHARED, dw->name, &dw->irq[0]); > > if (err) { > > dw->nr_irqs = 0; > > return err; > > } > > > > - get_cached_msi_msg(dw->ops->irq_vector(dev, 0), > > - &dw->irq[0].msi); > > + if (irq_get_msi_desc(irq)) > > + get_cached_msi_msg(irq, &dw->irq[0].msi); > > } else { > > /* Distribute IRQs equally among all channels */ > > int tmp = dw->nr_irqs; > > @@ -804,7 +806,8 @@ static int dw_edma_irq_request(struct dw_edma_chip *chip, > > dw_edma_add_irq_mask(&rd_mask, *rd_alloc, dw->rd_ch_cnt); > > > > for (i = 0; i < (*wr_alloc + *rd_alloc); i++) { > > - err = request_irq(dw->ops->irq_vector(dev, i), > > + irq = dw->ops->irq_vector(dev, i); > > + err = request_irq(irq, > > i < *wr_alloc ? > > dw_edma_interrupt_write : > > dw_edma_interrupt_read, > > @@ -815,8 +818,8 @@ static int dw_edma_irq_request(struct dw_edma_chip *chip, > > return err; > > } > > > > - get_cached_msi_msg(dw->ops->irq_vector(dev, i), > > - &dw->irq[i].msi); > > + if (irq_get_msi_desc(irq)) > > + get_cached_msi_msg(irq, &dw->irq[i].msi); > > } > > > > dw->nr_irqs = i; > > -- > > 2.7.4 > > > Acked-by: Gustavo Pimentel <gustavo.pimentel@xxxxxxxxxxxx> Thanks Gustavo for the Ack. FYI, I first considered adding an ops function to decouple dw-edma-core.c from struct msi_msg. However, in a separate use case that I have in mind, dw-edma would run on a host system having Synopsys DesignWare PCI eDMA hardware on the host-side. In this use case, the host system eDMA engines may be used in conjunction with an endpoint device also having the same eDMA hardware. In this use case, dw-edma running on the host would need to call get_cached_msi_msg() just in case the host has an msi_msg cached from the endpoint device. As a result, I opted to not add a new ops function. Regards, Alan > >
