From: Ralph Campbell <ralph.campbell-h88ZbnxC6KDQT0dZR+AlfA@public.gmane.org>
To: Roland Dreier <rdreier-FYB4Gu1CFyUAvxtiuMwx3w@public.gmane.org>
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Subject: [PATCH v3 49/52] IB/qib: Add qib_wc_x86_64.c
Date: Thu, 06 May 2010 17:03:09 -0700 [thread overview]
Message-ID: <20100507000309.3441.73777.stgit@chromite.mv.qlogic.com> (raw)
In-Reply-To: <20100506235849.3441.85930.stgit-/vjeY7uYZjrPXfVEPVhPGq6RkeBMCJyt@public.gmane.org>
creates the qib_wc_x86_64.c file.
Signed-off-by: Ralph Campbell <ralph.campbell-h88ZbnxC6KDQT0dZR+AlfA@public.gmane.org>
---
drivers/infiniband/hw/qib/qib_wc_x86_64.c | 171 +++++++++++++++++++++++++++++
1 files changed, 171 insertions(+), 0 deletions(-)
create mode 100644 drivers/infiniband/hw/qib/qib_wc_x86_64.c
diff --git a/drivers/infiniband/hw/qib/qib_wc_x86_64.c b/drivers/infiniband/hw/qib/qib_wc_x86_64.c
new file mode 100644
index 0000000..561b8bc
--- /dev/null
+++ b/drivers/infiniband/hw/qib/qib_wc_x86_64.c
@@ -0,0 +1,171 @@
+/*
+ * Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
+ * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+/*
+ * This file is conditionally built on x86_64 only. Otherwise weak symbol
+ * versions of the functions exported from here are used.
+ */
+
+#include <linux/pci.h>
+#include <asm/mtrr.h>
+#include <asm/processor.h>
+
+#include "qib.h"
+
+/**
+ * qib_enable_wc - enable write combining for MMIO writes to the device
+ * @dd: qlogic_ib device
+ *
+ * This routine is x86_64-specific; it twiddles the CPU's MTRRs to enable
+ * write combining.
+ */
+int qib_enable_wc(struct qib_devdata *dd)
+{
+ int ret = 0;
+ u64 pioaddr, piolen;
+ unsigned bits;
+ const unsigned long addr = pci_resource_start(dd->pcidev, 0);
+ const size_t len = pci_resource_len(dd->pcidev, 0);
+
+ /*
+ * Set the PIO buffers to be WCCOMB, so we get HT bursts to the
+ * chip. Linux (possibly the hardware) requires it to be on a power
+ * of 2 address matching the length (which has to be a power of 2).
+ * For rev1, that means the base address, for rev2, it will be just
+ * the PIO buffers themselves.
+ * For chips with two sets of buffers, the calculations are
+ * somewhat more complicated; we need to sum, and the piobufbase
+ * register has both offsets, 2K in low 32 bits, 4K in high 32 bits.
+ * The buffers are still packed, so a single range covers both.
+ */
+ if (dd->piobcnt2k && dd->piobcnt4k) {
+ /* 2 sizes for chip */
+ unsigned long pio2kbase, pio4kbase;
+ pio2kbase = dd->piobufbase & 0xffffffffUL;
+ pio4kbase = (dd->piobufbase >> 32) & 0xffffffffUL;
+ if (pio2kbase < pio4kbase) {
+ /* all current chips */
+ pioaddr = addr + pio2kbase;
+ piolen = pio4kbase - pio2kbase +
+ dd->piobcnt4k * dd->align4k;
+ } else {
+ pioaddr = addr + pio4kbase;
+ piolen = pio2kbase - pio4kbase +
+ dd->piobcnt2k * dd->palign;
+ }
+ } else { /* single buffer size (2K, currently) */
+ pioaddr = addr + dd->piobufbase;
+ piolen = dd->piobcnt2k * dd->palign +
+ dd->piobcnt4k * dd->align4k;
+ }
+
+ for (bits = 0; !(piolen & (1ULL << bits)); bits++)
+ /* do nothing */ ;
+
+ if (piolen != (1ULL << bits)) {
+ piolen >>= bits;
+ while (piolen >>= 1)
+ bits++;
+ piolen = 1ULL << (bits + 1);
+ }
+ if (pioaddr & (piolen - 1)) {
+ u64 atmp;
+ atmp = pioaddr & ~(piolen - 1);
+ if (atmp < addr || (atmp + piolen) > (addr + len)) {
+ qib_dev_err(dd, "No way to align address/size "
+ "(%llx/%llx), no WC mtrr\n",
+ (unsigned long long) atmp,
+ (unsigned long long) piolen << 1);
+ ret = -ENODEV;
+ } else {
+ pioaddr = atmp;
+ piolen <<= 1;
+ }
+ }
+
+ if (!ret) {
+ int cookie;
+
+ cookie = mtrr_add(pioaddr, piolen, MTRR_TYPE_WRCOMB, 0);
+ if (cookie < 0) {
+ {
+ qib_devinfo(dd->pcidev,
+ "mtrr_add() WC for PIO bufs "
+ "failed (%d)\n",
+ cookie);
+ ret = -EINVAL;
+ }
+ } else {
+ dd->wc_cookie = cookie;
+ dd->wc_base = (unsigned long) pioaddr;
+ dd->wc_len = (unsigned long) piolen;
+ }
+ }
+
+ return ret;
+}
+
+/**
+ * qib_disable_wc - disable write combining for MMIO writes to the device
+ * @dd: qlogic_ib device
+ */
+void qib_disable_wc(struct qib_devdata *dd)
+{
+ if (dd->wc_cookie) {
+ int r;
+
+ r = mtrr_del(dd->wc_cookie, dd->wc_base,
+ dd->wc_len);
+ if (r < 0)
+ qib_devinfo(dd->pcidev,
+ "mtrr_del(%lx, %lx, %lx) failed: %d\n",
+ dd->wc_cookie, dd->wc_base,
+ dd->wc_len, r);
+ dd->wc_cookie = 0; /* even on failure */
+ }
+}
+
+/**
+ * qib_unordered_wc - indicate whether write combining is ordered
+ *
+ * Because our performance depends on our ability to do write combining mmio
+ * writes in the most efficient way, we need to know if we are on an Intel
+ * or AMD x86_64 processor. AMD x86_64 processors flush WC buffers out in
+ * the order completed, and so no special flushing is required to get
+ * correct ordering. Intel processors, however, will flush write buffers
+ * out in "random" orders, and so explicit ordering is needed at times.
+ */
+int qib_unordered_wc(void)
+{
+ return boot_cpu_data.x86_vendor != X86_VENDOR_AMD;
+}
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2010-05-07 0:03 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-05-06 23:58 [PATCH 0/52] IB/qib: add Ralph Campbell
[not found] ` <20100506235849.3441.85930.stgit-/vjeY7uYZjrPXfVEPVhPGq6RkeBMCJyt@public.gmane.org>
2010-05-06 23:58 ` [PATCH v3 01/52] IB/qib: Add Kconfig Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 02/52] IB/qib: Add Makefile Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 03/52] IB/qib: Add qib.h Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 04/52] IB/qib: Add qib_6120_regs.h Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 05/52] IB/qib: Add qib_7220.h Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 06/52] IB/qib: Add qib_7220_regs.h Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 08/52] IB/qib: Add qib_common.h Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 09/52] IB/qib: Add qib_cq.c Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 10/52] IB/qib: Add qib_diag.c Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 11/52] IB/qib: Add qib_dma.c Ralph Campbell
2010-05-06 23:59 ` [PATCH v3 12/52] IB/qib: Add qib_driver.c Ralph Campbell
2010-05-07 0:00 ` [PATCH v3 13/52] IB/qib: Add qib_eeprom.c Ralph Campbell
2010-05-07 0:00 ` [PATCH v3 14/52] IB/qib: Add qib_file_ops.c Ralph Campbell
2010-05-07 0:00 ` [PATCH v3 15/52] IB/qib: Add qib_fs.c Ralph Campbell
2010-05-07 0:00 ` [PATCH v3 19/52] IB/qib: Add qib_init.c Ralph Campbell
2010-05-07 0:00 ` [PATCH v3 20/52] IB/qib: Add qib_intr.c Ralph Campbell
2010-05-07 0:00 ` [PATCH v3 21/52] IB/qib: Add qib_keys.c Ralph Campbell
2010-05-07 0:00 ` [PATCH v3 22/52] IB/qib: Add qib_mad.c Ralph Campbell
2010-05-07 0:00 ` [PATCH v3 23/52] IB/qib: Add qib_mad.h Ralph Campbell
2010-05-07 0:00 ` [PATCH v3 24/52] IB/qib: Add qib_mmap.c Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 25/52] IB/qib: Add qib_mr.c Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 26/52] IB/qib: Add qib_pcie.c Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 27/52] IB/qib: Add qib_pio_copy.c Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 28/52] IB/qib: Add qib_qp.c Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 29/52] IB/qib: Add qib_qsfp.c Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 30/52] IB/qib: Add qib_qsfp.h Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 31/52] IB/qib: Add qib_rc.c Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 32/52] IB/qib: Add qib_ruc.c Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 33/52] IB/qib: Add qib_sd7220.c Ralph Campbell
[not found] ` <20100507000145.3441.59808.stgit-/vjeY7uYZjrPXfVEPVhPGq6RkeBMCJyt@public.gmane.org>
2010-05-10 22:50 ` Roland Dreier
2010-05-07 0:01 ` [PATCH v3 34/52] IB/qib: Add qib_sd7220_img.c Ralph Campbell
2010-05-07 0:01 ` [PATCH v3 35/52] IB/qib: Add qib_sdma.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 36/52] IB/qib: Add qib_srq.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 37/52] IB/qib: Add qib_sysfs.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 38/52] IB/qib: Add qib_twsi.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 39/52] IB/qib: Add qib_tx.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 40/52] IB/qib: Add qib_uc.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 41/52] IB/qib: Add qib_ud.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 42/52] IB/qib: Add qib_user_pages.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 43/52] IB/qib: Add qib_user_sdma.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 44/52] IB/qib: Add qib_user_sdma.h Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 45/52] IB/qib: Add qib_verbs.c Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 46/52] IB/qib: Add qib_verbs.h Ralph Campbell
2010-05-07 0:02 ` [PATCH v3 47/52] IB/qib: Add qib_verbs_mcast.c Ralph Campbell
2010-05-07 0:03 ` [PATCH v3 48/52] IB/qib: Add qib_wc_ppc64.c Ralph Campbell
2010-05-07 0:03 ` Ralph Campbell [this message]
2010-05-07 0:03 ` [PATCH v3 50/52] IB/qib: Hooks for adding the QIB driver into the framework Ralph Campbell
2010-05-07 0:03 ` [PATCH v3 52/52] IB/core: allow HCAs to create IB port sysfs files Ralph Campbell
[not found] ` <20100507000026.3441.87902.stgit@chromite.mv.qlogic.com>
[not found] ` <20100507000026.3441.87902.stgit-/vjeY7uYZjrPXfVEPVhPGq6RkeBMCJyt@public.gmane.org>
2010-05-10 22:55 ` [PATCH v3 18/52] IB/qib: Add qib_iba7322.c Roland Dreier
[not found] ` <adad3x3xt6q.fsf-BjVyx320WGW9gfZ95n9DRSW4+XlvGpQz@public.gmane.org>
2010-05-11 15:49 ` [PATCH v3 18/52] IB/qib: Add qib_iba7322.c (serdes parameters) Dave Olson
[not found] ` <alpine.LFD.1.10.1005110837180.7212-vxnkQ4oxbxUi9g6yJnKVd0EOCMrvLtNR@public.gmane.org>
2010-05-11 21:55 ` Roland Dreier
[not found] ` <adad3x2w1c3.fsf-BjVyx320WGW9gfZ95n9DRSW4+XlvGpQz@public.gmane.org>
2010-05-12 17:22 ` Dave Olson
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20100507000309.3441.73777.stgit@chromite.mv.qlogic.com \
--to=ralph.campbell-h88zbnxc6kdqt0dzr+alfa@public.gmane.org \
--cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=rdreier-FYB4Gu1CFyUAvxtiuMwx3w@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox