[PATCH 3/8] iommu/iova: Add flush counters to Flush-Queue implementation

From: Joerg Roedel
Date: Fri Aug 11 2017 - 10:44:14 EST


From: Joerg Roedel <jroedel@xxxxxxx>

There are two counters:

* fq_flush_start_cnt - Increased when a TLB flush
is started.

* fq_flush_finish_cnt - Increased when a TLB flush
is finished.

The fq_flush_start_cnt is assigned to every Flush-Queue
entry on its creation. When freeing entries from the
Flush-Queue, the value in the entry is compared to the
fq_flush_finish_cnt. The entry can only be freed when its
value is less than the value of fq_flush_finish_cnt.

The reason for these counters it to take advantage of IOMMU
TLB flushes that happened on other CPUs. These already
flushed the TLB for Flush-Queue entries on other CPUs so
that they can already be freed without flushing the TLB
again.

This makes it less likely that the Flush-Queue is full and
saves IOMMU TLB flushes.

Signed-off-by: Joerg Roedel <jroedel@xxxxxxx>
---
drivers/iommu/iova.c | 27 ++++++++++++++++++++++++---
include/linux/iova.h | 8 ++++++++
2 files changed, 32 insertions(+), 3 deletions(-)

diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c
index e5c9a7a..47b144e 100644
--- a/drivers/iommu/iova.c
+++ b/drivers/iommu/iova.c
@@ -75,6 +75,9 @@ int init_iova_flush_queue(struct iova_domain *iovad,
{
int cpu;

+ atomic64_set(&iovad->fq_flush_start_cnt, 0);
+ atomic64_set(&iovad->fq_flush_finish_cnt, 0);
+
iovad->fq = alloc_percpu(struct iova_fq);
if (!iovad->fq)
return -ENOMEM;
@@ -482,20 +485,30 @@ static inline unsigned fq_ring_add(struct iova_fq *fq)

static void fq_ring_free(struct iova_domain *iovad, struct iova_fq *fq)
{
+ u64 counter = atomic64_read(&iovad->fq_flush_finish_cnt);
unsigned idx;

fq_ring_for_each(idx, fq) {

+ if (fq->entries[idx].counter >= counter)
+ break;
+
if (iovad->entry_dtor)
iovad->entry_dtor(fq->entries[idx].data);

free_iova_fast(iovad,
fq->entries[idx].iova_pfn,
fq->entries[idx].pages);
+
+ fq->head = (fq->head + 1) % IOVA_FQ_SIZE;
}
+}

- fq->head = 0;
- fq->tail = 0;
+static void iova_domain_flush(struct iova_domain *iovad)
+{
+ atomic64_inc(&iovad->fq_flush_start_cnt);
+ iovad->flush_cb(iovad);
+ atomic64_inc(&iovad->fq_flush_finish_cnt);
}

static void fq_destroy_all_entries(struct iova_domain *iovad)
@@ -526,8 +539,15 @@ void queue_iova(struct iova_domain *iovad,
struct iova_fq *fq = get_cpu_ptr(iovad->fq);
unsigned idx;

+ /*
+ * First remove all entries from the flush queue that have already been
+ * flushed out on another CPU. This makes the fq_full() check below less
+ * likely to be true.
+ */
+ fq_ring_free(iovad, fq);
+
if (fq_full(fq)) {
- iovad->flush_cb(iovad);
+ iova_domain_flush(iovad);
fq_ring_free(iovad, fq);
}

@@ -536,6 +556,7 @@ void queue_iova(struct iova_domain *iovad,
fq->entries[idx].iova_pfn = pfn;
fq->entries[idx].pages = pages;
fq->entries[idx].data = data;
+ fq->entries[idx].counter = atomic64_read(&iovad->fq_flush_start_cnt);

put_cpu_ptr(iovad->fq);
}
diff --git a/include/linux/iova.h b/include/linux/iova.h
index 1ae8524..985b800 100644
--- a/include/linux/iova.h
+++ b/include/linux/iova.h
@@ -14,6 +14,7 @@
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/rbtree.h>
+#include <linux/atomic.h>
#include <linux/dma-mapping.h>

/* iova structure */
@@ -52,6 +53,7 @@ struct iova_fq_entry {
unsigned long iova_pfn;
unsigned long pages;
unsigned long data;
+ u64 counter; /* Flush counter when this entrie was added */
};

/* Per-CPU Flush Queue structure */
@@ -77,6 +79,12 @@ struct iova_domain {
iova entry */

struct iova_fq __percpu *fq; /* Flush Queue */
+
+ atomic64_t fq_flush_start_cnt; /* Number of TLB flushes that
+ have been started */
+
+ atomic64_t fq_flush_finish_cnt; /* Number of TLB flushes that
+ have been finished */
};

static inline unsigned long iova_size(struct iova *iova)
--
2.7.4