summaryrefslogtreecommitdiff
path: root/drivers/infiniband/hw/hfi1/pcie.c
diff options
context:
space:
mode:
authorMike Marciniszyn <mike.marciniszyn@intel.com>2017-07-24 17:45:31 +0300
committerDoug Ledford <dledford@redhat.com>2017-07-31 22:17:54 +0300
commitcb51c5d2cda855302910ab352f3d391c1a00aba0 (patch)
tree87f909b33c8d08687af1c72aabeba04885921a86 /drivers/infiniband/hw/hfi1/pcie.c
parentc53df62c7a9a82c7375308af270f90a08e94f6b6 (diff)
downloadlinux-cb51c5d2cda855302910ab352f3d391c1a00aba0.tar.xz
IB/hfi1: Fix bar0 mapping to use write combining
When the debugpat kernel boot flag is turned on the following traces are printed: [ 1884.793168] x86/PAT: Overlap at 0x90000000-0x92000000 [ 1884.803510] x86/PAT: reserve_memtype added [mem 0x91200000-0x9127ffff], track uncached-minus, req write-combining, ret uncached-minus [ 1884.818167] hfi1 0000:05:00.0: hfi1_0: WC Remapped RcvArray: ffffc9000a980000 The ioremap_wc() clearly is not returning a write combining mapping due to an overlap where the RcvArray is mapped in a uncached mapping prior to creating the proposed write combining mapping. The patch replaces the single base register for uncached CSRs that used to overlap the RcvArray with two mappings. One, kregbase1, from the bar0 up to the RcvArray and another, kregbase2, from the end of the RcvArray to the pio send buffer space. A new dd field, base2_start, is used to convert the zero-based offset in the CSR routines to the correct kregbase1/kregbase2 mapping. A single direct write of the RcvArray CSRs is replaced with hfi1_put_tid() to insure correct access using the new disjoint mapping. Additionally, the kregend field is deleted since it is only ever written. patdebug now shows the RcvArray as write combining: [ 35.688990] x86/PAT: reserve_memtype added [mem 0x91200000-0x9127ffff], track write-combining, req write-combining, ret write-combining To insulate from any potential issues with write combining, all writeq are now flushed in hfi1_put_tid() and rcv_array_wc_fill(). Reviewed-by: Mitko Haralanov <mitko.haralanov@intel.com> Reviewed-by: Ashutosh Dixit <ashutosh.dixit@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Diffstat (limited to 'drivers/infiniband/hw/hfi1/pcie.c')
-rw-r--r--drivers/infiniband/hw/hfi1/pcie.c58
1 files changed, 42 insertions, 16 deletions
diff --git a/drivers/infiniband/hw/hfi1/pcie.c b/drivers/infiniband/hw/hfi1/pcie.c
index 903ea45bf650..cc7be224095d 100644
--- a/drivers/infiniband/hw/hfi1/pcie.c
+++ b/drivers/infiniband/hw/hfi1/pcie.c
@@ -180,31 +180,47 @@ int hfi1_pcie_ddinit(struct hfi1_devdata *dd, struct pci_dev *pdev)
return -EINVAL;
}
- dd->kregbase = ioremap_nocache(addr, TXE_PIO_SEND);
- if (!dd->kregbase)
+ dd->kregbase1 = ioremap_nocache(addr, RCV_ARRAY);
+ if (!dd->kregbase1) {
+ dd_dev_err(dd, "UC mapping of kregbase1 failed\n");
return -ENOMEM;
+ }
+ dd_dev_info(dd, "UC base1: %p for %x\n", dd->kregbase1, RCV_ARRAY);
+ dd->chip_rcv_array_count = readq(dd->kregbase1 + RCV_ARRAY_CNT);
+ dd_dev_info(dd, "RcvArray count: %u\n", dd->chip_rcv_array_count);
+ dd->base2_start = RCV_ARRAY + dd->chip_rcv_array_count * 8;
+
+ dd->kregbase2 = ioremap_nocache(
+ addr + dd->base2_start,
+ TXE_PIO_SEND - dd->base2_start);
+ if (!dd->kregbase2) {
+ dd_dev_err(dd, "UC mapping of kregbase2 failed\n");
+ goto nomem;
+ }
+ dd_dev_info(dd, "UC base2: %p for %x\n", dd->kregbase2,
+ TXE_PIO_SEND - dd->base2_start);
dd->piobase = ioremap_wc(addr + TXE_PIO_SEND, TXE_PIO_SIZE);
if (!dd->piobase) {
- iounmap(dd->kregbase);
- return -ENOMEM;
+ dd_dev_err(dd, "WC mapping of send buffers failed\n");
+ goto nomem;
}
+ dd_dev_info(dd, "WC piobase: %p\n for %x", dd->piobase, TXE_PIO_SIZE);
- dd->flags |= HFI1_PRESENT; /* now register routines work */
-
- dd->kregend = dd->kregbase + TXE_PIO_SEND;
dd->physaddr = addr; /* used for io_remap, etc. */
/*
- * Re-map the chip's RcvArray as write-combining to allow us
+ * Map the chip's RcvArray as write-combining to allow us
* to write an entire cacheline worth of entries in one shot.
- * If this re-map fails, just continue - the RcvArray programming
- * function will handle both cases.
*/
- dd->chip_rcv_array_count = read_csr(dd, RCV_ARRAY_CNT);
dd->rcvarray_wc = ioremap_wc(addr + RCV_ARRAY,
dd->chip_rcv_array_count * 8);
- dd_dev_info(dd, "WC Remapped RcvArray: %p\n", dd->rcvarray_wc);
+ if (!dd->rcvarray_wc) {
+ dd_dev_err(dd, "WC mapping of receive array failed\n");
+ goto nomem;
+ }
+ dd_dev_info(dd, "WC RcvArray: %p for %x\n",
+ dd->rcvarray_wc, dd->chip_rcv_array_count * 8);
/*
* Save BARs and command to rewrite after device reset.
*/
@@ -253,10 +269,16 @@ int hfi1_pcie_ddinit(struct hfi1_devdata *dd, struct pci_dev *pdev)
if (ret)
goto read_error;
+ dd->flags |= HFI1_PRESENT; /* chip.c CSR routines now work */
return 0;
read_error:
dd_dev_err(dd, "Unable to read from PCI config\n");
+ goto bail_error;
+nomem:
+ ret = -ENOMEM;
+bail_error:
+ hfi1_pcie_ddcleanup(dd);
return ret;
}
@@ -267,15 +289,19 @@ read_error:
*/
void hfi1_pcie_ddcleanup(struct hfi1_devdata *dd)
{
- u64 __iomem *base = (void __iomem *)dd->kregbase;
-
dd->flags &= ~HFI1_PRESENT;
- dd->kregbase = NULL;
- iounmap(base);
+ if (dd->kregbase1)
+ iounmap(dd->kregbase1);
+ dd->kregbase1 = NULL;
+ if (dd->kregbase2)
+ iounmap(dd->kregbase2);
+ dd->kregbase2 = NULL;
if (dd->rcvarray_wc)
iounmap(dd->rcvarray_wc);
+ dd->rcvarray_wc = NULL;
if (dd->piobase)
iounmap(dd->piobase);
+ dd->piobase = NULL;
}
/* return the PCIe link speed from the given link status */