All of lore.kernel.org
 help / color / mirror / Atom feed
From: muli@il.ibm.com
To: ak@suse.de
Cc: linux-kernel@vger.kernel.org, muli@il.ibm.com, jdmason@kudzu.us
Subject: [PATCH 2/4] x86-64:  Calgary - tighten up the bitmap locking
Date: Wed, 11 Jul 2007 16:35:12 +0300	[thread overview]
Message-ID: <11841609142004-git-send-email-muli@il.ibm.com> (raw)
In-Reply-To: <11841609141802-git-send-email-muli@il.ibm.com>

From: Muli Ben-Yehuda <muli@il.ibm.com>

Currently the IOMMU table's lock protects both the bitmap and access
to the hardware's TCE table. Access to the TCE table is synchronized
through the bitmap; therefore, only hold the lock while modifying the
bitmap. This gives a yummy 10-15% reduction in CPU utilization for
netperf on a large SMP machine.

Signed-off-by: Muli Ben-Yehuda <muli@il.ibm.com>
---
 arch/x86_64/kernel/pci-calgary.c |   40 ++++++++++++++++---------------------
 1 files changed, 17 insertions(+), 23 deletions(-)

diff --git a/arch/x86_64/kernel/pci-calgary.c b/arch/x86_64/kernel/pci-calgary.c
index f589c2d..28ce2b3 100644
--- a/arch/x86_64/kernel/pci-calgary.c
+++ b/arch/x86_64/kernel/pci-calgary.c
@@ -235,6 +235,7 @@ static void iommu_range_reserve(struct iommu_table *tbl,
 	unsigned long index;
 	unsigned long end;
 	unsigned long badbit;
+	unsigned long flags;
 
 	index = start_addr >> PAGE_SHIFT;
 
@@ -246,6 +247,8 @@ static void iommu_range_reserve(struct iommu_table *tbl,
 	if (end > tbl->it_size) /* don't go off the table */
 		end = tbl->it_size;
 
+	spin_lock_irqsave(&tbl->it_lock, flags);
+
 	badbit = verify_bit_range(tbl->it_map, 0, index, end);
 	if (badbit != ~0UL) {
 		if (printk_ratelimit())
@@ -255,15 +258,20 @@ static void iommu_range_reserve(struct iommu_table *tbl,
 	}
 
 	set_bit_string(tbl->it_map, index, npages);
+
+	spin_unlock_irqrestore(&tbl->it_lock, flags);
 }
 
 static unsigned long iommu_range_alloc(struct iommu_table *tbl,
 	unsigned int npages)
 {
+	unsigned long flags;
 	unsigned long offset;
 
 	BUG_ON(npages == 0);
 
+	spin_lock_irqsave(&tbl->it_lock, flags);
+
 	offset = find_next_zero_string(tbl->it_map, tbl->it_hint,
 				       tbl->it_size, npages);
 	if (offset == ~0UL) {
@@ -272,6 +280,7 @@ static unsigned long iommu_range_alloc(struct iommu_table *tbl,
 					       tbl->it_size, npages);
 		if (offset == ~0UL) {
 			printk(KERN_WARNING "Calgary: IOMMU full.\n");
+			spin_unlock_irqrestore(&tbl->it_lock, flags);
 			if (panic_on_overflow)
 				panic("Calgary: fix the allocator.\n");
 			else
@@ -283,17 +292,17 @@ static unsigned long iommu_range_alloc(struct iommu_table *tbl,
 	tbl->it_hint = offset + npages;
 	BUG_ON(tbl->it_hint > tbl->it_size);
 
+	spin_unlock_irqrestore(&tbl->it_lock, flags);
+
 	return offset;
 }
 
 static dma_addr_t iommu_alloc(struct iommu_table *tbl, void *vaddr,
 	unsigned int npages, int direction)
 {
-	unsigned long entry, flags;
+	unsigned long entry;
 	dma_addr_t ret = bad_dma_address;
 
-	spin_lock_irqsave(&tbl->it_lock, flags);
-
 	entry = iommu_range_alloc(tbl, npages);
 
 	if (unlikely(entry == bad_dma_address))
@@ -306,12 +315,9 @@ static dma_addr_t iommu_alloc(struct iommu_table *tbl, void *vaddr,
 	tce_build(tbl, entry, npages, (unsigned long)vaddr & PAGE_MASK,
 		  direction);
 
-	spin_unlock_irqrestore(&tbl->it_lock, flags);
-
 	return ret;
 
 error:
-	spin_unlock_irqrestore(&tbl->it_lock, flags);
 	printk(KERN_WARNING "Calgary: failed to allocate %u pages in "
 	       "iommu %p\n", npages, tbl);
 	return bad_dma_address;
@@ -323,6 +329,7 @@ static void __iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr,
 	unsigned long entry;
 	unsigned long badbit;
 	unsigned long badend;
+	unsigned long flags;
 
 	/* were we called with bad_dma_address? */
 	badend = bad_dma_address + (EMERGENCY_PAGES * PAGE_SIZE);
@@ -339,6 +346,8 @@ static void __iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr,
 
 	tce_free(tbl, entry, npages);
 
+	spin_lock_irqsave(&tbl->it_lock, flags);
+
 	badbit = verify_bit_range(tbl->it_map, 1, entry, entry + npages);
 	if (badbit != ~0UL) {
 		if (printk_ratelimit())
@@ -348,18 +357,14 @@ static void __iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr,
 	}
 
 	__clear_bit_string(tbl->it_map, entry, npages);
+
+	spin_unlock_irqrestore(&tbl->it_lock, flags);
 }
 
 static void iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr,
 	unsigned int npages)
 {
-	unsigned long flags;
-
-	spin_lock_irqsave(&tbl->it_lock, flags);
-
 	__iommu_free(tbl, dma_addr, npages);
-
-	spin_unlock_irqrestore(&tbl->it_lock, flags);
 }
 
 static inline struct iommu_table *find_iommu_table(struct device *dev)
@@ -404,17 +409,12 @@ static void __calgary_unmap_sg(struct iommu_table *tbl,
 void calgary_unmap_sg(struct device *dev, struct scatterlist *sglist,
 		      int nelems, int direction)
 {
-	unsigned long flags;
 	struct iommu_table *tbl = find_iommu_table(dev);
 
 	if (!translate_phb(to_pci_dev(dev)))
 		return;
 
-	spin_lock_irqsave(&tbl->it_lock, flags);
-
 	__calgary_unmap_sg(tbl, sglist, nelems, direction);
-
-	spin_unlock_irqrestore(&tbl->it_lock, flags);
 }
 
 static int calgary_nontranslate_map_sg(struct device* dev,
@@ -435,7 +435,6 @@ int calgary_map_sg(struct device *dev, struct scatterlist *sg,
 	int nelems, int direction)
 {
 	struct iommu_table *tbl = find_iommu_table(dev);
-	unsigned long flags;
 	unsigned long vaddr;
 	unsigned int npages;
 	unsigned long entry;
@@ -444,8 +443,6 @@ int calgary_map_sg(struct device *dev, struct scatterlist *sg,
 	if (!translate_phb(to_pci_dev(dev)))
 		return calgary_nontranslate_map_sg(dev, sg, nelems, direction);
 
-	spin_lock_irqsave(&tbl->it_lock, flags);
-
 	for (i = 0; i < nelems; i++ ) {
 		struct scatterlist *s = &sg[i];
 		BUG_ON(!s->page);
@@ -469,8 +466,6 @@ int calgary_map_sg(struct device *dev, struct scatterlist *sg,
 		s->dma_length = s->length;
 	}
 
-	spin_unlock_irqrestore(&tbl->it_lock, flags);
-
 	return nelems;
 error:
 	__calgary_unmap_sg(tbl, sg, nelems, direction);
@@ -478,7 +473,6 @@ error:
 		sg[i].dma_address = bad_dma_address;
 		sg[i].dma_length = 0;
 	}
-	spin_unlock_irqrestore(&tbl->it_lock, flags);
 	return 0;
 }
 
-- 
1.5.2


  reply	other threads:[~2007-07-11 13:36 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-07-11 13:35 Calgary: more updates for 2.6.23 muli
2007-07-11 13:35 ` [PATCH 1/4] x86-64: Calgary - fix few style problems pointed out by checkpatch.pl muli
2007-07-11 13:35   ` muli [this message]
2007-07-11 13:35     ` [PATCH 3/4] x86-64: Calgary - fold in redundant functions muli
2007-07-11 13:35       ` [PATCH 4/4] x86_64: Calgary - change _map_single, etc to static muli

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=11841609142004-git-send-email-muli@il.ibm.com \
    --to=muli@il.ibm.com \
    --cc=ak@suse.de \
    --cc=jdmason@kudzu.us \
    --cc=linux-kernel@vger.kernel.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 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.