iommu/iova: Retry from last rb tree node if iova search fails
Subsystems (2)
PATCH v2 1/2 iommu/iova: Retry from last rb tree node if iova search fails
From: Vijayanand Jitta <[email protected]>
When ever a new iova alloc request comes iova is always searched from the cached node and the nodes which are previous to cached node. So, even if there is free iova space available in the nodes which are next to the cached node iova allocation can still fail because of this approach.
Consider the following sequence of iova alloc and frees on 1GB of iova space
1) alloc - 500MB
2) alloc - 12MB
3) alloc - 499MB
4) free - 12MB which was allocated in step 2
5) alloc - 13MB
After the above sequence we will have 12MB of free iova space and cached node will be pointing to the iova pfn of last alloc of 13MB which will be the lowest iova pfn of that iova space. Now if we get an alloc request of 2MB we just search from cached node and then look for lower iova pfn's for free iova and as they aren't any, iova alloc fails though there is 12MB of free iova space.
To avoid such iova search failures do a retry from the last rb tree node when iova search fails, this will search the entire tree and get an iova if its available.
--- drivers/iommu/iova.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)
drivers/iommu/iova.c +17 -6
@@ -184,8 +184,9 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad,struct rb_node *curr, *prev;struct iova *curr_iova;unsigned long flags; -unsigned long new_pfn; +unsigned long new_pfn, low_pfn_new;unsigned long align_mask = ~0UL ; +unsigned long high_pfn = limit_pfn, low_pfn = iovad->start_pfn;if (size_aligned) align_mask <<=fls_long (size -1 );
@@ -198,15 +199,25 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad, curr =__get_cached_rbnode (iovad, limit_pfn); curr_iova = rb_entry(curr, struct iova, node); + low_pfn_new = curr_iova->pfn_hi +1 ; + +retry:do { - limit_pfn =min (limit_pfn, curr_iova->pfn_lo); - new_pfn = (limit_pfn - size) & align_mask; + high_pfn =min (high_pfn, curr_iova->pfn_lo); + new_pfn = (high_pfn - size) & align_mask; prev = curr; curr =rb_prev (curr); curr_iova = rb_entry(curr, struct iova, node); - }while (curr && new_pfn <= curr_iova->pfn_hi); - -if (limit_pfn < size || new_pfn < iovad->start_pfn) { + }while (curr && new_pfn <= curr_iova->pfn_hi && new_pfn >= low_pfn); + +if (high_pfn < size || new_pfn < low_pfn) { +if (low_pfn == iovad->start_pfn && low_pfn_new < limit_pfn) { + high_pfn = limit_pfn; + low_pfn = low_pfn_new; + curr = &iovad->anchor.node; + curr_iova = rb_entry(curr, struct iova, node); + goto retry; + } iovad->max32_alloc_size = size; goto iova32_full; }
Signature · 7 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation 1.9.1 _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
PATCH v2 2/2 iommu/iova: Free global iova rcache on iova alloc failure
From: Vijayanand Jitta <[email protected]>
When ever an iova alloc request fails we free the iova ranges present in the percpu iova rcaches and then retry but the global iova rcache is not freed as a result we could still see iova alloc failure even after retry as global rcache is holding the iova's which can cause fragmentation. So, free the global iova rcache as well and then go for the retry.
--- drivers/iommu/iova.c | 23 +++++++++++++++++++++++ include/linux/iova.h | 6 ++++++ 2 files changed, 29 insertions(+)
drivers/iommu/iova.c +23 -0
@@ -442,6 +442,7 @@ struct iova *find_iova(struct iova_domain *iovad, unsigned long pfn) flush_rcache = false;for_each_online_cpu (cpu)free_cpu_cached_iovas (cpu, iovad); +free_global_cached_iovas (iovad); goto retry; }
@@ -1055,5 +1056,27 @@ void free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad) } } +/* +* free all the IOVA ranges of global cache +*/ +void free_global_cached_iovas (struct iova_domain *iovad) +{ +struct iova_rcache *rcache; +unsigned long flags; +int i, j; + +for (i =0 ; i <IOVA_RANGE_CACHE_MAX_SIZE ; ++i) { + rcache = &iovad->rcaches[i]; +spin_lock_irqsave (&rcache->lock, flags); +for (j =0 ; j < rcache->depot_size; ++j) { +iova_magazine_free_pfns (rcache->depot[j], iovad); +iova_magazine_free (rcache->depot[j]); + rcache->depot[j] =NULL ; + } + rcache->depot_size =0 ; +spin_unlock_irqrestore (&rcache->lock, flags); + } +} +MODULE_AUTHOR ("Anil S Keshavamurthy <[email protected]>" );MODULE_LICENSE ("GPL" );
include/linux/iova.h +6 -0
@@ -163,6 +163,7 @@ int init_iova_flush_queue(struct iova_domain *iovad,struct iova *split_and_remove_iova (struct iova_domain *iovad,struct iova *iova,unsigned long pfn_lo,unsigned long pfn_hi);void free_cpu_cached_iovas (unsigned int cpu,struct iova_domain *iovad); +void free_global_cached_iovas (struct iova_domain *iovad);#else static inline int iova_cache_get (void ) {
@@ -270,6 +271,11 @@ static inline void free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad) { } + +static inline void free_global_cached_iovas (struct iova_domain *iovad) +{ +} +#endif #endif
Signature · 7 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation 1.9.1 _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
Robin Murphy
maintainerRe: [PATCH v2 2/2] iommu/iova: Free global iova rcache on iova alloc failure
Quoted 29 lines · On 2020-08-20 13:49, [email protected] wrote:
From: Vijayanand Jitta <[email protected]>
When ever an iova alloc request fails we free the iova ranges present in the percpu iova rcaches and then retry but the global iova rcache is not freed as a result we could still see iova alloc failure even after retry as global rcache is holding the iova's which can cause fragmentation. So, free the global iova rcache as well and then go for the retry.
Signed-off-by Vijayanand Jitta <[email protected]>--- drivers/iommu/iova.c | 23 +++++++++++++++++++++++ include/linux/iova.h | 6 ++++++ 2 files changed, 29 insertions(+)drivers/iommu/iova.c +1
-0@@ -442,6 +442,7 @@ struct iova *find_iova(struct iova_domain *iovad, unsigned long pfn) flush_rcache = false;for_each_online_cpu (cpu)free_cpu_cached_iovas (cpu, iovad); +free_global_cached_iovas (iovad); goto retry; }
drivers/iommu/iova.c void free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad) · line 1062
Quoted 7 lines
drivers/iommu/iova.c +4
-0@@ -1055,5 +1056,27 @@ void free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad) } } +/* +* free all the IOVA ranges of global cache +*/ +void free_global_cached_iovas (struct iova_domain *iovad)
As John pointed out last time, this should be static and the header changes dropped.
(TBH we should probably register our own hotplug notifier instance for a flush queue, so that external code has no need to poke at the per-CPU caches either)
Robin.
Quoted 45 lines
drivers/iommu/iova.c +18
-0+{ +struct iova_rcache *rcache; +unsigned long flags; +int i, j; + +for (i =0 ; i <IOVA_RANGE_CACHE_MAX_SIZE ; ++i) { + rcache = &iovad->rcaches[i]; +spin_lock_irqsave (&rcache->lock, flags); +for (j =0 ; j < rcache->depot_size; ++j) { +iova_magazine_free_pfns (rcache->depot[j], iovad); +iova_magazine_free (rcache->depot[j]); + rcache->depot[j] =NULL ; + } + rcache->depot_size =0 ; +spin_unlock_irqrestore (&rcache->lock, flags); + } +} +MODULE_AUTHOR ("Anil S Keshavamurthy <[email protected]>" );MODULE_LICENSE ("GPL" );include/linux/iova.h +6
-0@@ -163,6 +163,7 @@ int init_iova_flush_queue(struct iova_domain *iovad,struct iova *split_and_remove_iova (struct iova_domain *iovad,struct iova *iova,unsigned long pfn_lo,unsigned long pfn_hi);void free_cpu_cached_iovas (unsigned int cpu,struct iova_domain *iovad); +void free_global_cached_iovas (struct iova_domain *iovad);#else static inline int iova_cache_get (void ) {@@ -270,6 +271,11 @@ static inline void free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad) { } + +static inline void free_global_cached_iovas (struct iova_domain *iovad) +{ +} +#endif #endif
Re: [PATCH v2 2/2] iommu/iova: Free global iova rcache on iova alloc failure
Quoted 41 lines · Robin Murphy
Quoted 28 lines · On 2020-08-20 13:49, [email protected] wrote:
From: Vijayanand Jitta <[email protected]>
When ever an iova alloc request fails we free the iova ranges present in the percpu iova rcaches and then retry but the global iova rcache is not freed as a result we could still see iova alloc failure even after retry as global rcache is holding the iova's which can cause fragmentation. So, free the global iova rcache as well and then go for the retry.
Signed-off-by Vijayanand Jitta <[email protected]>---
drivers/iommu/iova.c| 23 +++++++++++++++++++++++include/linux/iova.h| 6 ++++++2 files changed, 29 insertions(+)drivers/iommu/iova.c +1
-0@@ -442,6 +442,7 @@ struct iova *find_iova(struct iova_domain *iovad, unsigned long pfn) flush_rcache = false; for_each_online_cpu(cpu) free_cpu_cached_iovas(cpu, iovad); + free_global_cached_iovas(iovad); goto retry; }Quoted 8 lines
@@ -1055,5 +1056,27 @@ void
free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad)
}
}
+/*
+ * free all the IOVA ranges of global cache
+ */
+voidfree_global_cached_iovas(struct iova_domain *iovad)As John pointed out last time, this should be static and the header changes dropped.
Quoted 6 lines
(TBH we should probably register our own hotplug notifier instance for a flush queue, so that external code has no need to poke at the per-CPU caches either)
Robin.
Right, I have made it static and dropped header changes in v3. can you please review that.
Thanks,
Vijay
Quoted 46 lines
Quoted 46 lines
drivers/iommu/iova.c +18
-0+{ + struct iova_rcache *rcache; + unsigned long flags; + int i, j; + + for (i = 0; i < IOVA_RANGE_CACHE_MAX_SIZE; ++i) { + rcache = &iovad->rcaches[i]; + spin_lock_irqsave(&rcache->lock, flags); + for (j = 0; j < rcache->depot_size; ++j) { + iova_magazine_free_pfns(rcache->depot[j], iovad); + iova_magazine_free(rcache->depot[j]); + rcache->depot[j] = NULL; + } + rcache->depot_size = 0; + spin_unlock_irqrestore(&rcache->lock, flags); + } +} +
MODULE_AUTHOR("Anil S Keshavamurthy <[email protected]>");MODULE_LICENSE("GPL");include/linux/iova.h +0
-0@@ -163,6 +163,7 @@ int init_iova_flush_queue(struct iova_domain *iovad, struct iova *split_and_remove_iova(struct iova_domain *iovad, struct iova *iova, unsigned long pfn_lo, unsigned long pfn_hi);void
free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad);
+voidfree_global_cached_iovas(struct iova_domain *iovad); #else
static inline intiova_cache_get(void)
{include/linux/iova.h +5
-0@@ -270,6 +271,11 @@ static inline void free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad) { } + +static inline void free_global_cached_iovas (struct iova_domain *iovad) +{ +} + #endif #endif
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation
Re: [PATCH v2 2/2] iommu/iova: Free global iova rcache on iova alloc failure
Quoted 48 lines · Vijayanand Jitta
Quoted 45 lines · Robin Murphy
Quoted 28 lines · On 2020-08-20 13:49, [email protected] wrote:
From: Vijayanand Jitta <[email protected]>
When ever an iova alloc request fails we free the iova ranges present in the percpu iova rcaches and then retry but the global iova rcache is not freed as a result we could still see iova alloc failure even after retry as global rcache is holding the iova's which can cause fragmentation. So, free the global iova rcache as well and then go for the retry.
Signed-off-by Vijayanand Jitta <[email protected]>---
drivers/iommu/iova.c| 23 +++++++++++++++++++++++include/linux/iova.h| 6 ++++++2 files changed, 29 insertions(+)drivers/iommu/iova.c +1
-0@@ -442,6 +442,7 @@ struct iova *find_iova(struct iova_domain *iovad, unsigned long pfn) flush_rcache = false; for_each_online_cpu(cpu) free_cpu_cached_iovas(cpu, iovad); + free_global_cached_iovas(iovad); goto retry; }Quoted 8 lines
@@ -1055,5 +1056,27 @@ void
free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad)
}
}
+/*
+ * free all the IOVA ranges of global cache
+ */
+voidfree_global_cached_iovas(struct iova_domain *iovad)As John pointed out last time, this should be static and the header changes dropped.
(TBH we should probably register our own hotplug notifier instance for a flush queue, so that external code has no need to poke at the per-CPU caches either)
Quoted 8 lines
Quoted 2 lines
Robin.
Right, I have made it static and dropped header changes in v3. can you please review that.
Thanks,
Vijay
Please review v4 instead of v3, I have updated other patch as well in v4.
Thanks,
Vijay
Quoted 47 lines
Quoted 46 lines
Quoted 46 lines
drivers/iommu/iova.c +18
-0+{ + struct iova_rcache *rcache; + unsigned long flags; + int i, j; + + for (i = 0; i < IOVA_RANGE_CACHE_MAX_SIZE; ++i) { + rcache = &iovad->rcaches[i]; + spin_lock_irqsave(&rcache->lock, flags); + for (j = 0; j < rcache->depot_size; ++j) { + iova_magazine_free_pfns(rcache->depot[j], iovad); + iova_magazine_free(rcache->depot[j]); + rcache->depot[j] = NULL; + } + rcache->depot_size = 0; + spin_unlock_irqrestore(&rcache->lock, flags); + } +} +
MODULE_AUTHOR("Anil S Keshavamurthy <[email protected]>");MODULE_LICENSE("GPL");include/linux/iova.h +0
-0@@ -163,6 +163,7 @@ int init_iova_flush_queue(struct iova_domain *iovad, struct iova *split_and_remove_iova(struct iova_domain *iovad, struct iova *iova, unsigned long pfn_lo, unsigned long pfn_hi);void
free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad);
+voidfree_global_cached_iovas(struct iova_domain *iovad); #else
static inline intiova_cache_get(void)
{include/linux/iova.h +5
-0@@ -270,6 +271,11 @@ static inline void free_cpu_cached_iovas(unsigned int cpu, struct iova_domain *iovad) { } + +static inline void free_global_cached_iovas (struct iova_domain *iovad) +{ +} + #endif #endif
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation
Quoted 70 lines · On 8/20/2020 6:19 PM, [email protected] wrote:
From: Vijayanand Jitta <[email protected]>
When ever a new iova alloc request comes iova is always searched from the cached node and the nodes which are previous to cached node. So, even if there is free iova space available in the nodes which are next to the cached node iova allocation can still fail because of this approach.
Consider the following sequence of iova alloc and frees on 1GB of iova space
1) alloc - 500MB
2) alloc - 12MB
3) alloc - 499MB
4) free - 12MB which was allocated in step 2
5) alloc - 13MBAfter the above sequence we will have 12MB of free iova space and cached node will be pointing to the iova pfn of last alloc of 13MB which will be the lowest iova pfn of that iova space. Now if we get an alloc request of 2MB we just search from cached node and then look for lower iova pfn's for free iova and as they aren't any, iova alloc fails though there is 12MB of free iova space.
To avoid such iova search failures do a retry from the last rb tree node when iova search fails, this will search the entire tree and get an iova if its available.
Signed-off-by Vijayanand Jitta <[email protected]>--- drivers/iommu/iova.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)drivers/iommu/iova.c +10
-6@@ -184,8 +184,9 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad,struct rb_node *curr, *prev;struct iova *curr_iova;unsigned long flags; -unsigned long new_pfn; +unsigned long new_pfn, low_pfn_new;unsigned long align_mask = ~0UL ; +unsigned long high_pfn = limit_pfn, low_pfn = iovad->start_pfn;if (size_aligned) align_mask <<=fls_long (size -1 );@@ -198,15 +199,25 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad, curr =__get_cached_rbnode (iovad, limit_pfn); curr_iova = rb_entry(curr, struct iova, node); + low_pfn_new = curr_iova->pfn_hi +1 ; + +retry:do { - limit_pfn =min (limit_pfn, curr_iova->pfn_lo); - new_pfn = (limit_pfn - size) & align_mask; + high_pfn =min (high_pfn, curr_iova->pfn_lo); + new_pfn = (high_pfn - size) & align_mask; prev = curr; curr =rb_prev (curr); curr_iova = rb_entry(curr, struct iova, node); - }while (curr && new_pfn <= curr_iova->pfn_hi); - -if (limit_pfn < size || new_pfn < iovad->start_pfn) { + }while (curr && new_pfn <= curr_iova->pfn_hi && new_pfn >= low_pfn); + +if (high_pfn < size || new_pfn < low_pfn) {
drivers/iommu/iova.c static int __alloc_and_insert_iova_range(struct iova_domain *iovad, · line 223
Quoted 10 lines
drivers/iommu/iova.c +7
-0+if (low_pfn == iovad->start_pfn && low_pfn_new < limit_pfn) { + high_pfn = limit_pfn; + low_pfn = low_pfn_new; + curr = &iovad->anchor.node; + curr_iova = rb_entry(curr, struct iova, node); + goto retry; + } iovad->max32_alloc_size = size; goto iova32_full; }
ping ?
Thanks,
Vijay
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation
Quoted 86 lines · Vijayanand Jitta
Quoted 70 lines · On 8/20/2020 6:19 PM, [email protected] wrote:
From: Vijayanand Jitta <[email protected]>
When ever a new iova alloc request comes iova is always searched from the cached node and the nodes which are previous to cached node. So, even if there is free iova space available in the nodes which are next to the cached node iova allocation can still fail because of this approach.
Consider the following sequence of iova alloc and frees on 1GB of iova space
1) alloc - 500MB
2) alloc - 12MB
3) alloc - 499MB
4) free - 12MB which was allocated in step 2
5) alloc - 13MBAfter the above sequence we will have 12MB of free iova space and cached node will be pointing to the iova pfn of last alloc of 13MB which will be the lowest iova pfn of that iova space. Now if we get an alloc request of 2MB we just search from cached node and then look for lower iova pfn's for free iova and as they aren't any, iova alloc fails though there is 12MB of free iova space.
To avoid such iova search failures do a retry from the last rb tree node when iova search fails, this will search the entire tree and get an iova if its available.
Signed-off-by Vijayanand Jitta <[email protected]>--- drivers/iommu/iova.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)drivers/iommu/iova.c +10
-6@@ -184,8 +184,9 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad,struct rb_node *curr, *prev;struct iova *curr_iova;unsigned long flags; -unsigned long new_pfn; +unsigned long new_pfn, low_pfn_new;unsigned long align_mask = ~0UL ; +unsigned long high_pfn = limit_pfn, low_pfn = iovad->start_pfn;if (size_aligned) align_mask <<=fls_long (size -1 );@@ -198,15 +199,25 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad, curr =__get_cached_rbnode (iovad, limit_pfn); curr_iova = rb_entry(curr, struct iova, node); + low_pfn_new = curr_iova->pfn_hi +1 ; + +retry:do { - limit_pfn =min (limit_pfn, curr_iova->pfn_lo); - new_pfn = (limit_pfn - size) & align_mask; + high_pfn =min (high_pfn, curr_iova->pfn_lo); + new_pfn = (high_pfn - size) & align_mask; prev = curr; curr =rb_prev (curr); curr_iova = rb_entry(curr, struct iova, node); - }while (curr && new_pfn <= curr_iova->pfn_hi); - -if (limit_pfn < size || new_pfn < iovad->start_pfn) { + }while (curr && new_pfn <= curr_iova->pfn_hi && new_pfn >= low_pfn); + +if (high_pfn < size || new_pfn < low_pfn) {drivers/iommu/iova.c static int __alloc_and_insert_iova_range(struct iova_domain *iovad, · line 223
Quoted 10 lines
drivers/iommu/iova.c +7
-0+if (low_pfn == iovad->start_pfn && low_pfn_new < limit_pfn) { + high_pfn = limit_pfn; + low_pfn = low_pfn_new; + curr = &iovad->anchor.node; + curr_iova = rb_entry(curr, struct iova, node); + goto retry; + } iovad->max32_alloc_size = size; goto iova32_full; }ping ?
Quoted 3 lines
Thanks,
Vijay
ping ?
Thanks,
Vijay
Signature · 6 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
Joerg Roedel
maintainerQuoted 1 line · Vijayanand Jitta
ping ?
Robin needs to have a look first.
Regards,
Joerg
Quoted 49 lines · On 2020-08-20 13:49, [email protected] wrote:
From: Vijayanand Jitta <[email protected]>
When ever a new iova alloc request comes iova is always searched from the cached node and the nodes which are previous to cached node. So, even if there is free iova space available in the nodes which are next to the cached node iova allocation can still fail because of this approach.
Consider the following sequence of iova alloc and frees on 1GB of iova space
1) alloc - 500MB
2) alloc - 12MB
3) alloc - 499MB
4) free - 12MB which was allocated in step 2
5) alloc - 13MBAfter the above sequence we will have 12MB of free iova space and cached node will be pointing to the iova pfn of last alloc of 13MB which will be the lowest iova pfn of that iova space. Now if we get an alloc request of 2MB we just search from cached node and then look for lower iova pfn's for free iova and as they aren't any, iova alloc fails though there is 12MB of free iova space.
To avoid such iova search failures do a retry from the last rb tree node when iova search fails, this will search the entire tree and get an iova if its available.
Signed-off-by Vijayanand Jitta <[email protected]>--- drivers/iommu/iova.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)drivers/iommu/iova.c +2
-1@@ -184,8 +184,9 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad,struct rb_node *curr, *prev;struct iova *curr_iova;unsigned long flags; -unsigned long new_pfn; +unsigned long new_pfn, low_pfn_new;unsigned long align_mask = ~0UL ; +unsigned long high_pfn = limit_pfn, low_pfn = iovad->start_pfn;if (size_aligned) align_mask <<=fls_long (size -1 );
drivers/iommu/iova.c static int __alloc_and_insert_iova_range(struct iova_domain *iovad, · line 202
Quoted 4 lines
drivers/iommu/iova.c +1
-0@@ -198,15 +199,25 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad, curr =__get_cached_rbnode (iovad, limit_pfn); curr_iova = rb_entry(curr, struct iova, node); + low_pfn_new = curr_iova->pfn_hi +1 ;
Could we call "low_pfn_new" something like "retry_pfn" instead? This code already has unavoidable readability struggles with so many different "pfn"s in play, so having two different meanings of "new" really doesn't help.
Other than that, I think this looks OK (IIRC it's basically what I originally suggested), so with the naming tweaked,
Quoted 27 lines
drivers/iommu/iova.c +14
-5+ +retry:do { - limit_pfn =min (limit_pfn, curr_iova->pfn_lo); - new_pfn = (limit_pfn - size) & align_mask; + high_pfn =min (high_pfn, curr_iova->pfn_lo); + new_pfn = (high_pfn - size) & align_mask; prev = curr; curr =rb_prev (curr); curr_iova = rb_entry(curr, struct iova, node); - }while (curr && new_pfn <= curr_iova->pfn_hi); - -if (limit_pfn < size || new_pfn < iovad->start_pfn) { + }while (curr && new_pfn <= curr_iova->pfn_hi && new_pfn >= low_pfn); + +if (high_pfn < size || new_pfn < low_pfn) { +if (low_pfn == iovad->start_pfn && low_pfn_new < limit_pfn) { + high_pfn = limit_pfn; + low_pfn = low_pfn_new; + curr = &iovad->anchor.node; + curr_iova = rb_entry(curr, struct iova, node); + goto retry; + } iovad->max32_alloc_size = size; goto iova32_full; }
Quoted 60 lines · Robin Murphy
Quoted 49 lines · On 2020-08-20 13:49, [email protected] wrote:
From: Vijayanand Jitta <[email protected]>
When ever a new iova alloc request comes iova is always searched from the cached node and the nodes which are previous to cached node. So, even if there is free iova space available in the nodes which are next to the cached node iova allocation can still fail because of this approach.
Consider the following sequence of iova alloc and frees on 1GB of iova space
1) alloc - 500MB
2) alloc - 12MB
3) alloc - 499MB
4) free - 12MB which was allocated in step 2
5) alloc - 13MBAfter the above sequence we will have 12MB of free iova space and cached node will be pointing to the iova pfn of last alloc of 13MB which will be the lowest iova pfn of that iova space. Now if we get an alloc request of 2MB we just search from cached node and then look for lower iova pfn's for free iova and as they aren't any, iova alloc fails though there is 12MB of free iova space.
To avoid such iova search failures do a retry from the last rb tree node when iova search fails, this will search the entire tree and get an iova if its available.
Signed-off-by Vijayanand Jitta <[email protected]>---
drivers/iommu/iova.c| 23 +++++++++++++++++------1 file changed, 17 insertions(+), 6 deletions(-)drivers/iommu/iova.c +2
-1@@ -184,8 +184,9 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad, struct rb_node *curr, *prev; struct iova *curr_iova; unsigned long flags; - unsigned long new_pfn; + unsigned long new_pfn, low_pfn_new; unsigned long align_mask = ~0UL; + unsigned long high_pfn = limit_pfn, low_pfn = iovad->start_pfn; if (size_aligned) align_mask <<= fls_long(size - 1);drivers/iommu/iova.c static int __alloc_and_insert_iova_range(struct · line 202
Quoted 4 lines
drivers/iommu/iova.c +1
-0@@ -198,15 +199,25 @@ static int __alloc_and_insert_iova_range(structiova_domain *iovad, curr = __get_cached_rbnode(iovad, limit_pfn); curr_iova = rb_entry(curr, struct iova, node); + low_pfn_new = curr_iova->pfn_hi + 1;Could we call "
low_pfn_new" something like "retry_pfn" instead? This code already has unavoidable readability struggles with so many different "pfn"s in play, so having two different meanings of "new" really doesn't help.
Quoted 5 lines
Other than that, I think this looks OK (IIRC it's basically what I originally suggested), so with the naming tweaked,
Reviewed-by Robin Murphy <[email protected]>
Thanks for review, I have renamed it to retry_pfn in v4.
Thanks,
Vijay
Quoted 28 lines
Quoted 28 lines
drivers/iommu/iova.c +14
-5+ +retry: do { - limit_pfn = min(limit_pfn, curr_iova->pfn_lo); - new_pfn = (limit_pfn - size) & align_mask; + high_pfn = min(high_pfn, curr_iova->pfn_lo); + new_pfn = (high_pfn - size) & align_mask; prev = curr; curr = rb_prev(curr); curr_iova = rb_entry(curr, struct iova, node); - } while (curr && new_pfn <= curr_iova->pfn_hi); - - if (limit_pfn < size || new_pfn < iovad->start_pfn) { + } while (curr && new_pfn <= curr_iova->pfn_hi && new_pfn >= low_pfn); + + if (high_pfn < size || new_pfn < low_pfn) { + if (low_pfn == iovad->start_pfn && low_pfn_new < limit_pfn) { + high_pfn = limit_pfn; + low_pfn = low_pfn_new; + curr = &iovad->anchor.node; + curr_iova = rb_entry(curr, struct iova, node); + goto retry; + } iovad->max32_alloc_size = size; goto iova32_full;}
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation