From: KyongHo Cho <pullip.cho@samsung.com>
To: KyongHo Cho <pullip.cho@samsung.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>,
Kukjin Kim <kgene.kim@samsung.com>,
Inho Lee <ilho215.lee@samsung.com>,
Inki Dae <inki.dae@samsung.com>,
Andrew Morton <akpm@linux-foundation.org>,
Ankita Garg <ankita@in.ibm.com>,
Daniel Walker <dwalker@codeaurora.org>,
Johan MOSSBERG <johan.xx.mossberg@stericsson.com>,
KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>,
Marek Szyprowski <m.szyprowski@samsung.com>,
Mel Gorman <mel@csn.ul.ie>,
linux-arm-kernel@lists.infradead.org,
linux-kernel@vger.kernel.org, linux-media@vger.kernel.org,
linux-mm@kvack.org, linux-samsung-soc@vger.kernel.org
Subject: [RFCv2,2/8] mm: vcm: reference counting on a reservation added
Date: Fri, 17 Dec 2010 12:56:21 +0900 [thread overview]
Message-ID: <1292558187-17348-3-git-send-email-pullip.cho@samsung.com> (raw)
In-Reply-To: <1292558187-17348-2-git-send-email-pullip.cho@samsung.com>
This commits adds vcm_ref_reserve() and refcnt member into vcm_res
structure. This feature is enabled by turnning on
CONFIG_VCM_RES_REFCNT. This enables the users of the vcm framework
not to care about the sequence of reserving and unreserving
in complex scenarios.
Signed-off-by: KyongHo Cho <pullip.cho@samsung.com>
---
Documentation/virtual-contiguous-memory.txt | 47 +++++++++++++++++++++++++++
include/linux/vcm.h | 23 +++++++++++++
mm/Kconfig | 7 ++++
mm/vcm.c | 17 ++++++++++
4 files changed, 94 insertions(+), 0 deletions(-)
diff --git a/Documentation/virtual-contiguous-memory.txt b/Documentation/virtual-contiguous-memory.txt
index 9793a86..2008465 100644
--- a/Documentation/virtual-contiguous-memory.txt
+++ b/Documentation/virtual-contiguous-memory.txt
@@ -275,6 +275,34 @@ To deactivate the VCM context vcm_deactivate() function is used:
Both of those functions can be called several times if all calls to
vcm_activate() are paired with a later call to vcm_deactivate().
+** Aquiring and releasing ownership of a reservation
+
+Once a device driver reserve a reservation, it may want to pass other device
+drivers or attach the reservation in a data structre. Since the reservation
+may be shared among many device drivers, the VCM context is needed to provide
+a simple way to unreserve a reservation.
+
+Below 2 functions gives the ownership of a reservation to the caller:
+
+ struct vcm_res *__must_check
+ vcm_reserve(struct vcm *vcm, resource_size_t size, unsigned flags);
+
+ int __must_check vcm_ref_reserve(struct vcm_res *res);
+
+vcm_reserve() creates a new reservation, thus the first owner of the
+reservation is set to the caller of vcm_reservation(). It then passes to a
+function the reservation. The function that received the reservation calls
+vcm_ref_reserve() to acquire the ownership of the given reservation. If the
+function decides that it does not need the reservation any more, it calls
+vcm_release() to release the ownership of the reservation.
+
+ void vcm_unreserve(struct vcm_res *res);
+
+It is not required to determine if other functions and drivers still need to
+access the reservation because this function just release the ownership of the
+reservation. If vcm_unreserve() finds no one has the ownership of the given
+reservation, only then does it unreserve (remove) the given reservation.
+
** Device driver example
The following is a simple, untested example of how platform and
@@ -706,6 +734,25 @@ automatically reflect new mappings on the hardware MMU.
Neither of the operations are required and if missing, VCM will
assume they are a no-operation and no warning will be generated.
+*** Ownership of a reservation
+
+When to aquire the ownership of a reservation:
+ - When creating a new reservation (having the ownership automatically)
+ - When assigning a reservation to a member of data structure
+ - When a reservation is passed from the caller function
+ - When requiring to access a reservation (that is a global variable)
+ at first in its context.
+The first one is done with vcm_reserve() and others with vcm_ref_reserve().
+
+When to release the ownership of a reservation:
+ - When a reservation is no longer needed in its context.
+ - When returning from a function and the function received a reservation
+ from its caller and acquired the ownership of the reservation.
+ - When removing a reservation from a data structure that includes a pointer
+ to the reservation
+It is not required as well unable to remove the reservation explicitly. The
+last call to vcm_unreserve() will cause the reservation to be removed.
+
* Epilogue
The initial version of the VCM framework was written by Zach Pfeffer
diff --git a/include/linux/vcm.h b/include/linux/vcm.h
index 965dc9b..3d54f18 100644
--- a/include/linux/vcm.h
+++ b/include/linux/vcm.h
@@ -52,6 +52,9 @@ struct vcm {
* @bound_size: number of bytes actually bound to the virtual address;
* read only.
* @res_size: size of the reserved address space in bytes; read only.
+ * @refcnt: reference count of a reservation to pass ownership of
+ * a reservation in a safe way; internal.
+ * Implemented only when CONFIG_VCM_RES_REFCNT is enabled.
* @vcm: VCM context; internal, read only for MMU drivers.
* @phys: pointer to physical memory bound to this reservation; NULL
* if no physical memory is bound; read only.
@@ -65,6 +68,9 @@ struct vcm_res {
dma_addr_t start;
resource_size_t bound_size;
resource_size_t res_size;
+#ifdef CONFIG_VCM_RES_REFCNT
+ atomic_t refcnt;
+#endif
struct vcm *vcm;
struct vcm_phys *phys;
@@ -180,6 +186,23 @@ struct vcm_res *__must_check
vcm_reserve(struct vcm *vcm, resource_size_t size, unsigned flags);
/**
+ * vcm_ref_reserve() - acquires the ownership of a reservation.
+ * @res: a valid reservation to access
+ *
+ * On success returns 0 and leads the same effect as vcm_reserve() in the
+ * context of the caller of this function. In other words, once a function
+ * acquire the ownership of a reservation with vcm_ref_reserve(), it must
+ * release the ownership with vcm_release() as soon as it does not need
+ * the reservation.
+ *
+ * On error returns -EINVAL. The only reason of the error is passing an invalid
+ * reservation like NULL or an unreserved reservation.
+ */
+#ifdef CONFIG_VCM_RES_REFCNT
+int __must_check vcm_ref_reserve(struct vcm_res *res);
+#endif
+
+/**
* vcm_unreserve() - destroyers a virtual address space reservation
* @res: reservation to destroy.
*
diff --git a/mm/Kconfig b/mm/Kconfig
index 7f0e4b1..b937f32 100644
--- a/mm/Kconfig
+++ b/mm/Kconfig
@@ -353,6 +353,13 @@ config VCM
<Documentation/virtual-contiguous-memory.txt>. If unsure, say
"n".
+config VCM_RES_REFCNT
+ bool "Reference counting on reservations"
+ depends on VCM
+ help
+ This enables reference counting on a reservation to make sharing
+ and migrating the ownership of the reservation easier.
+
#
# UP and nommu archs use km based percpu allocator
#
diff --git a/mm/vcm.c b/mm/vcm.c
index 1389ee6..5819f0f 100644
--- a/mm/vcm.c
+++ b/mm/vcm.c
@@ -156,10 +156,23 @@ vcm_reserve(struct vcm *vcm, resource_size_t size, unsigned flags)
__vcm_alloc_and_reserve(vcm, size, NULL, 0, &res, flags);
+#ifdef CONFIG_VCM_RES_REFCNT
+ if (!IS_ERR(res))
+ atomic_inc(&res->refcnt);
+#endif
+
return res;
}
EXPORT_SYMBOL_GPL(vcm_reserve);
+int __must_check vcm_ref_reserve(struct vcm_res *res)
+{
+ if (WARN_ON(!res) || (atomic_inc_return(&res->refcnt) < 2))
+ return -EINVAL;
+ return 0;
+}
+EXPORT_SYMBOL_GPL(vcm_ref_reserve);
+
struct vcm_res *__must_check
vcm_map(struct vcm *vcm, struct vcm_phys *phys, unsigned flags)
{
@@ -196,6 +209,10 @@ EXPORT_SYMBOL_GPL(vcm_map);
void vcm_unreserve(struct vcm_res *res)
{
if (!WARN_ON(!res)) {
+#ifdef CONFIG_VCM_RES_REFCNT
+ if (!atomic_dec_and_test(&res->refcnt))
+ return;
+#endif
if (WARN_ON(res->phys))
vcm_unbind(res);
if (!WARN_ON_ONCE(!res->vcm->driver->unreserve))
--
1.6.2.5
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom policy in Canada: sign http://dissolvethecrtc.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
next prev parent reply other threads:[~2010-12-17 4:25 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-12-17 3:56 [RFCv2,0/8] mm: vcm: The Virtual Memory Manager for multiple IOMMUs KyongHo Cho
2010-12-17 3:56 ` [RFCv2,1/8] mm: vcm: Virtual Contiguous Memory framework added KyongHo Cho
2010-12-17 3:56 ` KyongHo Cho [this message]
2010-12-17 3:56 ` [RFCv2,3/8] mm: vcm: physical memory allocator added KyongHo Cho
2010-12-17 3:56 ` [RFCv2,4/8] mm: vcm: VCM VMM driver added KyongHo Cho
2010-12-17 3:56 ` [RFCv2,5/8] mm: vcm: VCM MMU wrapper added KyongHo Cho
2010-12-17 3:56 ` [RFCv2,6/8] mm: vcm: VCM One-to-One " KyongHo Cho
2010-12-17 3:56 ` [RFCv2,7/8] mm: vcm: vcm-cma: VCM CMA driver added KyongHo Cho
2010-12-17 3:56 ` [RFCv2,8/8] mm: vcm: Sample " KyongHo Cho
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=1292558187-17348-3-git-send-email-pullip.cho@samsung.com \
--to=pullip.cho@samsung.com \
--cc=akpm@linux-foundation.org \
--cc=ankita@in.ibm.com \
--cc=dwalker@codeaurora.org \
--cc=ilho215.lee@samsung.com \
--cc=inki.dae@samsung.com \
--cc=johan.xx.mossberg@stericsson.com \
--cc=kamezawa.hiroyu@jp.fujitsu.com \
--cc=kgene.kim@samsung.com \
--cc=kyungmin.park@samsung.com \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-media@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux-samsung-soc@vger.kernel.org \
--cc=m.szyprowski@samsung.com \
--cc=mel@csn.ul.ie \
/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