From: Thomas Klein <tklein@de.ibm.com>
To: Jeff Garzik <jeff@garzik.org>
Cc: badari@us.ibm.com, Jan-Bernd Themann <themann@de.ibm.com>,
netdev <netdev@vger.kernel.org>,
tollefso@us.ibm.com, Hannes Hering <hering2@de.ibm.com>,
haveblue@us.ibm.com, linux-kernel <linux-kernel@vger.kernel.org>,
linux-ppc <linuxppc-dev@ozlabs.org>,
Christoph Raisch <raisch@de.ibm.com>
Subject: [PATCH] ehea: Add hugepage detection
Date: Mon, 27 Oct 2008 10:38:46 +0100 [thread overview]
Message-ID: <200810271038.46825.tklein@de.ibm.com> (raw)
All kernel memory which is used for kernel/hardware data transfer must be registered
with firmware using "memory regions". 16GB hugepages may not be part of a memory region
due to firmware restrictions.
This patch modifies the walk_memory_resource callback fn to filter hugepages and add
only standard memory to the busmap which is later on used for MR registration.
Signed-off-by: Thomas Klein <tklein@de.ibm.com>
---
This reworked patch accounts for comments I got. It's based on davem's net-2.6.git.
diff -Nurp -X dontdiff linux-2.6.27/drivers/net/ehea/ehea.h patched_kernel/drivers/net/ehea/ehea.h
--- linux-2.6.27/drivers/net/ehea/ehea.h 2008-10-24 09:29:19.000000000 +0200
+++ patched_kernel/drivers/net/ehea/ehea.h 2008-10-27 09:27:46.000000000 +0100
@@ -40,7 +40,7 @@
#include <asm/io.h>
#define DRV_NAME "ehea"
-#define DRV_VERSION "EHEA_0094"
+#define DRV_VERSION "EHEA_0095"
/* eHEA capability flags */
#define DLPAR_PORT_ADD_REM 1
diff -Nurp -X dontdiff linux-2.6.27/drivers/net/ehea/ehea_qmr.c patched_kernel/drivers/net/ehea/ehea_qmr.c
--- linux-2.6.27/drivers/net/ehea/ehea_qmr.c 2008-10-24 09:29:19.000000000 +0200
+++ patched_kernel/drivers/net/ehea/ehea_qmr.c 2008-10-27 09:27:46.000000000 +0100
@@ -632,10 +632,13 @@ static void ehea_rebuild_busmap(void)
}
}
-static int ehea_update_busmap(unsigned long pfn, unsigned long pgnum, int add)
+static int ehea_update_busmap(unsigned long pfn, unsigned long nr_pages, int add)
{
unsigned long i, start_section, end_section;
+ if (!nr_pages)
+ return 0;
+
if (!ehea_bmap) {
ehea_bmap = kzalloc(sizeof(struct ehea_bmap), GFP_KERNEL);
if (!ehea_bmap)
@@ -643,7 +646,7 @@ static int ehea_update_busmap(unsigned l
}
start_section = (pfn * PAGE_SIZE) / EHEA_SECTSIZE;
- end_section = start_section + ((pgnum * PAGE_SIZE) / EHEA_SECTSIZE);
+ end_section = start_section + ((nr_pages * PAGE_SIZE) / EHEA_SECTSIZE);
/* Mark entries as valid or invalid only; address is assigned later */
for (i = start_section; i < end_section; i++) {
u64 flag;
@@ -692,10 +695,54 @@ int ehea_rem_sect_bmap(unsigned long pfn
return ret;
}
-static int ehea_create_busmap_callback(unsigned long pfn,
- unsigned long nr_pages, void *arg)
+static int ehea_is_hugepage(unsigned long pfn)
+{
+ int page_order;
+
+ if (pfn & EHEA_HUGEPAGE_PFN_MASK)
+ return 0;
+
+ page_order = compound_order(pfn_to_page(pfn));
+ if (page_order + PAGE_SHIFT != EHEA_HUGEPAGESHIFT)
+ return 0;
+
+ return 1;
+}
+
+static int ehea_create_busmap_callback(unsigned long initial_pfn,
+ unsigned long total_nr_pages, void *arg)
{
- return ehea_update_busmap(pfn, nr_pages, EHEA_BUSMAP_ADD_SECT);
+ int ret;
+ unsigned long pfn, start_pfn, end_pfn, nr_pages;
+
+ if ((total_nr_pages * PAGE_SIZE) < EHEA_HUGEPAGE_SIZE)
+ return ehea_update_busmap(initial_pfn, total_nr_pages,
+ EHEA_BUSMAP_ADD_SECT);
+
+ /* Given chunk is >= 16GB -> check for hugepages */
+ start_pfn = initial_pfn;
+ end_pfn = initial_pfn + total_nr_pages;
+ pfn = start_pfn;
+
+ while (pfn < end_pfn) {
+ if (ehea_is_hugepage(pfn)) {
+ /* Add mem found in front of the hugepage */
+ nr_pages = pfn - start_pfn;
+ ret = ehea_update_busmap(start_pfn, nr_pages,
+ EHEA_BUSMAP_ADD_SECT);
+ if (ret)
+ return ret;
+
+ /* Skip the hugepage */
+ pfn += (EHEA_HUGEPAGE_SIZE / PAGE_SIZE);
+ start_pfn = pfn;
+ } else
+ pfn += (EHEA_SECTSIZE / PAGE_SIZE);
+ }
+
+ /* Add mem found behind the hugepage(s) */
+ nr_pages = pfn - start_pfn;
+ return ehea_update_busmap(start_pfn, nr_pages, EHEA_BUSMAP_ADD_SECT);
}
int ehea_create_busmap(void)
diff -Nurp -X dontdiff linux-2.6.27/drivers/net/ehea/ehea_qmr.h patched_kernel/drivers/net/ehea/ehea_qmr.h
--- linux-2.6.27/drivers/net/ehea/ehea_qmr.h 2008-10-24 09:29:19.000000000 +0200
+++ patched_kernel/drivers/net/ehea/ehea_qmr.h 2008-10-27 09:27:46.000000000 +0100
@@ -40,6 +40,9 @@
#define EHEA_PAGESIZE (1UL << EHEA_PAGESHIFT)
#define EHEA_SECTSIZE (1UL << 24)
#define EHEA_PAGES_PER_SECTION (EHEA_SECTSIZE >> EHEA_PAGESHIFT)
+#define EHEA_HUGEPAGESHIFT 34
+#define EHEA_HUGEPAGE_SIZE (1UL << EHEA_HUGEPAGESHIFT)
+#define EHEA_HUGEPAGE_PFN_MASK ((EHEA_HUGEPAGE_SIZE - 1) >> PAGE_SHIFT)
#if ((1UL << SECTION_SIZE_BITS) < EHEA_SECTSIZE)
#error eHEA module cannot work if kernel sectionsize < ehea sectionsize
WARNING: multiple messages have this Message-ID (diff)
From: Thomas Klein <tklein@de.ibm.com>
To: Jeff Garzik <jeff@garzik.org>
Cc: Christoph Raisch <raisch@de.ibm.com>,
Hannes Hering <hering2@de.ibm.com>,
"Jan-Bernd Themann" <themann@de.ibm.com>,
"linux-kernel" <linux-kernel@vger.kernel.org>,
"linux-ppc" <linuxppc-dev@ozlabs.org>,
"netdev" <netdev@vger.kernel.org>,
badari@us.ibm.com, haveblue@us.ibm.com, tollefso@us.ibm.com
Subject: [PATCH] ehea: Add hugepage detection
Date: Mon, 27 Oct 2008 10:38:46 +0100 [thread overview]
Message-ID: <200810271038.46825.tklein@de.ibm.com> (raw)
All kernel memory which is used for kernel/hardware data transfer must be registered
with firmware using "memory regions". 16GB hugepages may not be part of a memory region
due to firmware restrictions.
This patch modifies the walk_memory_resource callback fn to filter hugepages and add
only standard memory to the busmap which is later on used for MR registration.
Signed-off-by: Thomas Klein <tklein@de.ibm.com>
---
This reworked patch accounts for comments I got. It's based on davem's net-2.6.git.
diff -Nurp -X dontdiff linux-2.6.27/drivers/net/ehea/ehea.h patched_kernel/drivers/net/ehea/ehea.h
--- linux-2.6.27/drivers/net/ehea/ehea.h 2008-10-24 09:29:19.000000000 +0200
+++ patched_kernel/drivers/net/ehea/ehea.h 2008-10-27 09:27:46.000000000 +0100
@@ -40,7 +40,7 @@
#include <asm/io.h>
#define DRV_NAME "ehea"
-#define DRV_VERSION "EHEA_0094"
+#define DRV_VERSION "EHEA_0095"
/* eHEA capability flags */
#define DLPAR_PORT_ADD_REM 1
diff -Nurp -X dontdiff linux-2.6.27/drivers/net/ehea/ehea_qmr.c patched_kernel/drivers/net/ehea/ehea_qmr.c
--- linux-2.6.27/drivers/net/ehea/ehea_qmr.c 2008-10-24 09:29:19.000000000 +0200
+++ patched_kernel/drivers/net/ehea/ehea_qmr.c 2008-10-27 09:27:46.000000000 +0100
@@ -632,10 +632,13 @@ static void ehea_rebuild_busmap(void)
}
}
-static int ehea_update_busmap(unsigned long pfn, unsigned long pgnum, int add)
+static int ehea_update_busmap(unsigned long pfn, unsigned long nr_pages, int add)
{
unsigned long i, start_section, end_section;
+ if (!nr_pages)
+ return 0;
+
if (!ehea_bmap) {
ehea_bmap = kzalloc(sizeof(struct ehea_bmap), GFP_KERNEL);
if (!ehea_bmap)
@@ -643,7 +646,7 @@ static int ehea_update_busmap(unsigned l
}
start_section = (pfn * PAGE_SIZE) / EHEA_SECTSIZE;
- end_section = start_section + ((pgnum * PAGE_SIZE) / EHEA_SECTSIZE);
+ end_section = start_section + ((nr_pages * PAGE_SIZE) / EHEA_SECTSIZE);
/* Mark entries as valid or invalid only; address is assigned later */
for (i = start_section; i < end_section; i++) {
u64 flag;
@@ -692,10 +695,54 @@ int ehea_rem_sect_bmap(unsigned long pfn
return ret;
}
-static int ehea_create_busmap_callback(unsigned long pfn,
- unsigned long nr_pages, void *arg)
+static int ehea_is_hugepage(unsigned long pfn)
+{
+ int page_order;
+
+ if (pfn & EHEA_HUGEPAGE_PFN_MASK)
+ return 0;
+
+ page_order = compound_order(pfn_to_page(pfn));
+ if (page_order + PAGE_SHIFT != EHEA_HUGEPAGESHIFT)
+ return 0;
+
+ return 1;
+}
+
+static int ehea_create_busmap_callback(unsigned long initial_pfn,
+ unsigned long total_nr_pages, void *arg)
{
- return ehea_update_busmap(pfn, nr_pages, EHEA_BUSMAP_ADD_SECT);
+ int ret;
+ unsigned long pfn, start_pfn, end_pfn, nr_pages;
+
+ if ((total_nr_pages * PAGE_SIZE) < EHEA_HUGEPAGE_SIZE)
+ return ehea_update_busmap(initial_pfn, total_nr_pages,
+ EHEA_BUSMAP_ADD_SECT);
+
+ /* Given chunk is >= 16GB -> check for hugepages */
+ start_pfn = initial_pfn;
+ end_pfn = initial_pfn + total_nr_pages;
+ pfn = start_pfn;
+
+ while (pfn < end_pfn) {
+ if (ehea_is_hugepage(pfn)) {
+ /* Add mem found in front of the hugepage */
+ nr_pages = pfn - start_pfn;
+ ret = ehea_update_busmap(start_pfn, nr_pages,
+ EHEA_BUSMAP_ADD_SECT);
+ if (ret)
+ return ret;
+
+ /* Skip the hugepage */
+ pfn += (EHEA_HUGEPAGE_SIZE / PAGE_SIZE);
+ start_pfn = pfn;
+ } else
+ pfn += (EHEA_SECTSIZE / PAGE_SIZE);
+ }
+
+ /* Add mem found behind the hugepage(s) */
+ nr_pages = pfn - start_pfn;
+ return ehea_update_busmap(start_pfn, nr_pages, EHEA_BUSMAP_ADD_SECT);
}
int ehea_create_busmap(void)
diff -Nurp -X dontdiff linux-2.6.27/drivers/net/ehea/ehea_qmr.h patched_kernel/drivers/net/ehea/ehea_qmr.h
--- linux-2.6.27/drivers/net/ehea/ehea_qmr.h 2008-10-24 09:29:19.000000000 +0200
+++ patched_kernel/drivers/net/ehea/ehea_qmr.h 2008-10-27 09:27:46.000000000 +0100
@@ -40,6 +40,9 @@
#define EHEA_PAGESIZE (1UL << EHEA_PAGESHIFT)
#define EHEA_SECTSIZE (1UL << 24)
#define EHEA_PAGES_PER_SECTION (EHEA_SECTSIZE >> EHEA_PAGESHIFT)
+#define EHEA_HUGEPAGESHIFT 34
+#define EHEA_HUGEPAGE_SIZE (1UL << EHEA_HUGEPAGESHIFT)
+#define EHEA_HUGEPAGE_PFN_MASK ((EHEA_HUGEPAGE_SIZE - 1) >> PAGE_SHIFT)
#if ((1UL << SECTION_SIZE_BITS) < EHEA_SECTSIZE)
#error eHEA module cannot work if kernel sectionsize < ehea sectionsize
next reply other threads:[~2008-10-27 9:39 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-10-27 9:38 Thomas Klein [this message]
2008-10-27 9:38 ` [PATCH] ehea: Add hugepage detection Thomas Klein
2008-10-27 18:52 ` Jeff Garzik
2008-10-27 18:52 ` Jeff Garzik
-- strict thread matches above, loose matches on Subject: below --
2008-10-27 9:33 Thomas Klein
2008-10-27 9:33 ` Thomas Klein
2008-10-24 11:07 Thomas Klein
2008-10-24 18:55 ` Dave Hansen
2008-10-24 18:55 ` Dave Hansen
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=200810271038.46825.tklein@de.ibm.com \
--to=tklein@de.ibm.com \
--cc=badari@us.ibm.com \
--cc=haveblue@us.ibm.com \
--cc=hering2@de.ibm.com \
--cc=jeff@garzik.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linuxppc-dev@ozlabs.org \
--cc=netdev@vger.kernel.org \
--cc=raisch@de.ibm.com \
--cc=themann@de.ibm.com \
--cc=tollefso@us.ibm.com \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.