Re: [PATCH 02/53] IB/qib: Add Makefile (ipath vs qib)
On Mon, 23 Nov 2009, Roland Dreier wrote: | | +ib_qib-$(CONFIG_PCI_MSI) += qib_iba6120.o | | | | Seems you put the 6120 in the module device table even if PCI_MSI isn't | | set... does the driver build without MSI enabled? | | Yes. It just doesn't work well. It works well enough to tell people | what the problem is, which is what we wanted. | | No, it doesn't even build with CONFIG_PCI_MSI=n (and spews a bunch of | warnings before failing). Seriously, try it. We'll fix it. It was working; we missed the breakage. | The ipath driver has some significant problems with verbs, and those | problems have been fixed based on the significant restructuring and | rewrites in the ib_qib driver, so it's unlikely that any customer would | want to continue to use the ib_ipath driver in any case. | | I don't think we can drop 6110 support so blithely. I just did a quick | check, and brand-new HTX HCAs still seem readily available on ebay, so I They are likely available because they are old and slow, and few people can use them (because so few HTX-capable systems were ever built). If you feel strongly enough about it, ipath can remain as is, we just don't plan any work on it (including stripping out other board support, although of course others are welcome to do so). Dave Olson dave.ol...@qlogic.com -- To unsubscribe from this list: send the line unsubscribe linux-rdma in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
[PATCH 02/53] IB/qib: Add Makefile
This creates the Makefile file. Signed-off-by: Ralph Campbell ralph.campb...@qlogic.com --- drivers/infiniband/hw/qib/Makefile | 46 1 files changed, 46 insertions(+), 0 deletions(-) create mode 100644 drivers/infiniband/hw/qib/Makefile diff --git a/drivers/infiniband/hw/qib/Makefile b/drivers/infiniband/hw/qib/Makefile new file mode 100644 index 000..5dc5134 --- /dev/null +++ b/drivers/infiniband/hw/qib/Makefile @@ -0,0 +1,46 @@ +ccflags-y += -DQIB_KERN_TYPE=0 -DQIB_IDSTR='QLogic kernel.org driver' + +obj-$(CONFIG_INFINIBAND_QIB) += ib_qib.o + +ib_qib-y := \ + qib_cq.o \ + qib_diag.o \ + qib_dma.o \ + qib_driver.o \ + qib_eeprom.o \ + qib_file_ops.o \ + qib_fs.o \ + qib_init.o \ + qib_intr.o \ + qib_keys.o \ + qib_mad.o \ + qib_mmap.o \ + qib_mr.o \ + qib_pcie.o \ + qib_pio_copy.o \ + qib_qp.o \ + qib_qsfp.o \ + qib_rc.o \ + qib_ruc.o \ + qib_sdma.o \ + qib_srq.o \ + qib_sysfs.o \ + qib_trace.o \ + qib_twsi.o \ + qib_tx.o \ + qib_uc.o \ + qib_ud.o \ + qib_user_pages.o \ + qib_user_sdma.o \ + qib_verbs_mcast.o \ + qib_iba7220.o \ + qib_sd7220.o \ + qib_sd7220_img.o \ + qib_iba7322.o \ + qib_verbs.o + +# 6120 has no fallback if no MSI interrupts, others can do INTx +ib_qib-$(CONFIG_PCI_MSI) += qib_iba6120.o + +ib_qib-$(CONFIG_X86_64) += qib_wc_x86_64.o +ib_qib-$(CONFIG_PPC64) += qib_wc_ppc64.o -- To unsubscribe from this list: send the line unsubscribe linux-rdma in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 02/53] IB/qib: Add Makefile
+ccflags-y += -DQIB_KERN_TYPE=0 -DQIB_IDSTR='QLogic kernel.org driver' Do we really need to put this in the Makefile? Why can't this be done in some header file? What the heck is the KERN_TYPE anyway? +ib_qib-y := \ +qib_cq.o \ +qib_diag.o \ +qib_dma.o \ +qib_driver.o \ +qib_eeprom.o \ +qib_file_ops.o \ +qib_fs.o \ +qib_init.o \ +qib_intr.o \ +qib_keys.o \ I think the convention is to do this with multiple files per line, ie ib_qib-y := qip_cq.o qib_diag.o qib_dma.o ... \ ... +ib_qib-$(CONFIG_PCI_MSI) += qib_iba6120.o Seems you put the 6120 in the module device table even if PCI_MSI isn't set... does the driver build without MSI enabled? And are you dropping 6110 support (I see no HT_IRQ stuff)? So do we need to keep ipath around for that? - R. -- To unsubscribe from this list: send the line unsubscribe linux-rdma in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 02/53] IB/qib: Add Makefile
On Thu, 19 Nov 2009, Roland Dreier wrote: | | +ccflags-y += -DQIB_KERN_TYPE=0 -DQIB_IDSTR='QLogic kernel.org driver' | | Do we really need to put this in the Makefile? Why can't this be done | in some header file? What the heck is the KERN_TYPE anyway? A way of distinguishing qlogic-built drivers from elsewhere. We'll fix this; it shouldn't be needed in the upstream driver. | | I think the convention is to do this with multiple files per line, ie | | ib_qib-y := qip_cq.o qib_diag.o qib_dma.o ... \ Last time we did that, we got asked to change it to one per line. If conventions have changed, no problem. | +ib_qib-$(CONFIG_PCI_MSI) += qib_iba6120.o | | Seems you put the 6120 in the module device table even if PCI_MSI isn't | set... does the driver build without MSI enabled? Yes. It just doesn't work well. It works well enough to tell people what the problem is, which is what we wanted. | And are you dropping 6110 support (I see no HT_IRQ stuff)? So do we | need to keep ipath around for that? 6110 is out of this driver, and people (customers) are aware that support has been dropped. The current (non-verbs) userland software won't work on those boards any longer, and customers were using them primarily for non-verbs MPI. The ipath driver has some significant problems with verbs, and those problems have been fixed based on the significant restructuring and rewrites in the ib_qib driver, so it's unlikely that any customer would want to continue to use the ib_ipath driver in any case. It seems to me that people who want the ib_ipath driver are likely to use it only on older kernels where it was known to work OK, in any case. Testing on ib_ipath (by QLogic, at least) stopped with the 2.6.24 kernel. Dave Olson dave.ol...@qlogic.com -- To unsubscribe from this list: send the line unsubscribe linux-rdma in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html