From a18501074055879ef1d93a91ad4f1e95238f3e76 Mon Sep 17 00:00:00 2001 From: Zhang Zekun Date: Sun, 10 Dec 2023 14:10:30 +0800 Subject: [PATCH 1/6] Revert "iommu/iova: move IOVA_MAX_GLOBAL_MAGS outside of IOMMU_SUPPORT" hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8KS9A CVE: NA ------------------------------------------- This reverts commit b86a67670a8c3e5ce85237807f21490eba495b6b. Signed-off-by: Zhang Zekun (cherry picked from commit 74b94ca15330e8b09fc5d117588f4a210c081cbd) --- drivers/iommu/Kconfig | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/iommu/Kconfig b/drivers/iommu/Kconfig index b630e58c49b6..54d4a8cc3876 100644 --- a/drivers/iommu/Kconfig +++ b/drivers/iommu/Kconfig @@ -437,9 +437,6 @@ config SMMU_BYPASS_DEV This feature will be replaced by ACPI IORT RMR node, which will be upstreamed in mainline. - -endif # IOMMU_SUPPORT - config IOVA_MAX_GLOBAL_MAGS int "Set the max iova global magzines in iova rcache" range 16 2048 @@ -449,3 +446,6 @@ config IOVA_MAX_GLOBAL_MAGS it can be a bottle neck when lots of cpus are contending to use it. If you are suffering from the speed of allocing iova with more than 128 cpus, try to tune this config larger. + + +endif # IOMMU_SUPPORT -- Gitee From 8044083a2eb839938547f5b6552ce9a3a52c098f Mon Sep 17 00:00:00 2001 From: Zhang Zekun Date: Sun, 10 Dec 2023 14:10:31 +0800 Subject: [PATCH 2/6] Revert "config: enable set the max iova mag size to 128" hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8KS9A CVE: NA --------------------------------------------- This reverts commit ae2040f5837dacf454ed427109adc9123a3235ae. Signed-off-by: Zhang Zekun (cherry picked from commit 6ca1b8598aef38915286ae724aa85f778da9c190) --- arch/arm64/configs/openeuler_defconfig | 1 - 1 file changed, 1 deletion(-) diff --git a/arch/arm64/configs/openeuler_defconfig b/arch/arm64/configs/openeuler_defconfig index e82181af14d6..9021c1180637 100644 --- a/arch/arm64/configs/openeuler_defconfig +++ b/arch/arm64/configs/openeuler_defconfig @@ -5957,7 +5957,6 @@ CONFIG_ARM_SMMU_V3_PM=y # CONFIG_QCOM_IOMMU is not set # CONFIG_VIRTIO_IOMMU is not set CONFIG_SMMU_BYPASS_DEV=y -CONFIG_IOVA_MAX_GLOBAL_MAGS=128 # # Remoteproc drivers -- Gitee From 23a7b1e6317f637d5a55353f7590fbc097e6ac78 Mon Sep 17 00:00:00 2001 From: Zhang Zekun Date: Sun, 10 Dec 2023 14:10:32 +0800 Subject: [PATCH 3/6] Revert "iommu/iova: increase the iova_rcache depot max size to 128" hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8KS9A CVE: NA --------------------------------------- This reverts commit 0227a749c263db30ef5d49fc009a5161ae70cfab. Signed-off-by: Zhang Zekun (cherry picked from commit 1d51572e7cd5d8db9b4af13dbd5af3f4b47a656d) --- drivers/iommu/Kconfig | 10 ---------- include/linux/iova.h | 2 +- 2 files changed, 1 insertion(+), 11 deletions(-) diff --git a/drivers/iommu/Kconfig b/drivers/iommu/Kconfig index 54d4a8cc3876..f04a2bde0018 100644 --- a/drivers/iommu/Kconfig +++ b/drivers/iommu/Kconfig @@ -437,15 +437,5 @@ config SMMU_BYPASS_DEV This feature will be replaced by ACPI IORT RMR node, which will be upstreamed in mainline. -config IOVA_MAX_GLOBAL_MAGS - int "Set the max iova global magzines in iova rcache" - range 16 2048 - default "32" - help - Iova rcache global magizine is shared among every cpu. The size of - it can be a bottle neck when lots of cpus are contending to use it. - If you are suffering from the speed of allocing iova with more than - 128 cpus, try to tune this config larger. - endif # IOMMU_SUPPORT diff --git a/include/linux/iova.h b/include/linux/iova.h index 3cb469b366d7..dfa51ae49666 100644 --- a/include/linux/iova.h +++ b/include/linux/iova.h @@ -26,7 +26,7 @@ struct iova_magazine; struct iova_cpu_rcache; #define IOVA_RANGE_CACHE_MAX_SIZE 6 /* log of max cached IOVA range size (in pages) */ -#define MAX_GLOBAL_MAGS CONFIG_IOVA_MAX_GLOBAL_MAGS /* magazines per bin */ +#define MAX_GLOBAL_MAGS 32 /* magazines per bin */ struct iova_rcache { spinlock_t lock; -- Gitee From bac3143f4451426ea41713f8e4a69dccb5cc287d Mon Sep 17 00:00:00 2001 From: Feng Tang Date: Sun, 10 Dec 2023 14:10:33 +0800 Subject: [PATCH 4/6] iommu/iova: change IOVA_MAG_SIZE to 127 to save memory mainline inclusion from mainline-v5.19-rc5 commit b4c9bf178ace26c52c9cac36f265ba95132a221d bugzilla: https://gitee.com/openeuler/kernel/issues/I8KS9A CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b4c9bf178ace26c52c9cac36f265ba95132a221d ------------------------------------------- kmalloc will round up the request size to power of 2, and current iova_magazine's size is 1032 (1024+8) bytes, so each instance allocated will get 2048 bytes from kmalloc, causing around 1KB waste. Change IOVA_MAG_SIZE from 128 to 127 to make size of 'iova_magazine' 1024 bytes so that no memory will be wasted. Signed-off-by: Feng Tang Acked-by: Robin Murphy Link: https://lore.kernel.org/r/20220703114450.15184-1-feng.tang@intel.com Signed-off-by: Joerg Roedel Signed-off-by: Zhang Zekun (cherry picked from commit 26164f200d26e7f21def010a2f315539544c0fef) --- drivers/iommu/iova.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c index 1246e8f8bf08..b6d317b28e10 100644 --- a/drivers/iommu/iova.c +++ b/drivers/iommu/iova.c @@ -785,7 +785,12 @@ split_and_remove_iova(struct iova_domain *iovad, struct iova *iova, * dynamic size tuning described in the paper. */ -#define IOVA_MAG_SIZE 128 +/* + * As kmalloc's buffer size is fixed to power of 2, 127 is chosen to + * assure size of 'iova_magazine' to be 1024 bytes, so that no memory + * will be wasted. + */ +#define IOVA_MAG_SIZE 127 struct iova_magazine { unsigned long size; -- Gitee From 57ee49e17c0ccb96c2cf75cd28a4f2a473bcfaeb Mon Sep 17 00:00:00 2001 From: Zhang Zekun Date: Sun, 10 Dec 2023 14:10:34 +0800 Subject: [PATCH 5/6] iommu/iova: Make the rcache depot scale better mainline inclusion from mainline-v6.7-rc1 commit 911aa1245da83ff5e76d33bb612d8b5a3f2ec4a5 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8KS9A CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=911aa1245da83ff5e76d33bb612d8b5a3f2ec4a5 -------------------------------------- The algorithm in the original paper specifies the storage of full magazines in the depot as an unbounded list rather than a fixed-size array. It turns out to be pretty straightforward to do this in our implementation with no significant loss of efficiency. This allows the depot to scale up to the working set sizes of larger systems, while also potentially saving some memory on smaller ones too. Since this involves touching struct iova_magazine with the requisite care, we may as well reinforce the comment with a proper assertion too. Reviewed-by: John Garry Reviewed-by: Jerry Snitselaar Signed-off-by: Robin Murphy Link: https://lore.kernel.org/r/f597aa72fc3e1d315bc4574af0ce0ebe5c31cd22.1694535580.git.robin.murphy@arm.com Signed-off-by: Joerg Roedel Signed-off-by: Zhang Zekun (cherry picked from commit 876b598ef137b53bf85122571e20c4b6ab11e931) --- drivers/iommu/iova.c | 48 ++++++++++++++++++++++++++------------------ include/linux/iova.h | 4 +--- 2 files changed, 29 insertions(+), 23 deletions(-) diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c index b6d317b28e10..09ee0307f9f9 100644 --- a/drivers/iommu/iova.c +++ b/drivers/iommu/iova.c @@ -788,14 +788,19 @@ split_and_remove_iova(struct iova_domain *iovad, struct iova *iova, /* * As kmalloc's buffer size is fixed to power of 2, 127 is chosen to * assure size of 'iova_magazine' to be 1024 bytes, so that no memory - * will be wasted. + * will be wasted. Since only full magazines are inserted into the depot, + * we don't need to waste PFN capacity on a separate list head either. */ #define IOVA_MAG_SIZE 127 struct iova_magazine { - unsigned long size; + union { + unsigned long size; + struct iova_magazine *next; + }; unsigned long pfns[IOVA_MAG_SIZE]; }; +static_assert(!(sizeof(struct iova_magazine) & (sizeof(struct iova_magazine) - 1))); struct iova_cpu_rcache { spinlock_t lock; @@ -875,6 +880,21 @@ static void iova_magazine_push(struct iova_magazine *mag, unsigned long pfn) mag->pfns[mag->size++] = pfn; } +static struct iova_magazine *iova_depot_pop(struct iova_rcache *rcache) +{ + struct iova_magazine *mag = rcache->depot; + + rcache->depot = mag->next; + mag->size = IOVA_MAG_SIZE; + return mag; +} + +static void iova_depot_push(struct iova_rcache *rcache, struct iova_magazine *mag) +{ + mag->next = rcache->depot; + rcache->depot = mag; +} + static void init_iova_rcaches(struct iova_domain *iovad) { struct iova_cpu_rcache *cpu_rcache; @@ -885,7 +905,6 @@ static void init_iova_rcaches(struct iova_domain *iovad) for (i = 0; i < IOVA_RANGE_CACHE_MAX_SIZE; ++i) { rcache = &iovad->rcaches[i]; spin_lock_init(&rcache->lock); - rcache->depot_size = 0; rcache->cpu_rcaches = __alloc_percpu(sizeof(*cpu_rcache), cache_line_size()); if (WARN_ON(!rcache->cpu_rcaches)) continue; @@ -908,7 +927,6 @@ static bool __iova_rcache_insert(struct iova_domain *iovad, struct iova_rcache *rcache, unsigned long iova_pfn) { - struct iova_magazine *mag_to_free = NULL; struct iova_cpu_rcache *cpu_rcache; bool can_insert = false; unsigned long flags; @@ -926,12 +944,7 @@ static bool __iova_rcache_insert(struct iova_domain *iovad, if (new_mag) { spin_lock(&rcache->lock); - if (rcache->depot_size < MAX_GLOBAL_MAGS) { - rcache->depot[rcache->depot_size++] = - cpu_rcache->loaded; - } else { - mag_to_free = cpu_rcache->loaded; - } + iova_depot_push(rcache, cpu_rcache->loaded); spin_unlock(&rcache->lock); cpu_rcache->loaded = new_mag; @@ -944,11 +957,6 @@ static bool __iova_rcache_insert(struct iova_domain *iovad, spin_unlock_irqrestore(&cpu_rcache->lock, flags); - if (mag_to_free) { - iova_magazine_free_pfns(mag_to_free, iovad); - iova_magazine_free(mag_to_free); - } - return can_insert; } @@ -986,9 +994,9 @@ static unsigned long __iova_rcache_get(struct iova_rcache *rcache, has_pfn = true; } else { spin_lock(&rcache->lock); - if (rcache->depot_size > 0) { + if (rcache->depot) { iova_magazine_free(cpu_rcache->loaded); - cpu_rcache->loaded = rcache->depot[--rcache->depot_size]; + cpu_rcache->loaded = iova_depot_pop(rcache); has_pfn = true; } spin_unlock(&rcache->lock); @@ -1027,7 +1035,7 @@ static void free_iova_rcaches(struct iova_domain *iovad) struct iova_rcache *rcache; struct iova_cpu_rcache *cpu_rcache; unsigned int cpu; - int i, j; + int i; for (i = 0; i < IOVA_RANGE_CACHE_MAX_SIZE; ++i) { rcache = &iovad->rcaches[i]; @@ -1037,8 +1045,8 @@ static void free_iova_rcaches(struct iova_domain *iovad) iova_magazine_free(cpu_rcache->prev); } free_percpu(rcache->cpu_rcaches); - for (j = 0; j < rcache->depot_size; ++j) - iova_magazine_free(rcache->depot[j]); + while (rcache->depot) + iova_magazine_free(iova_depot_pop(rcache)); } } diff --git a/include/linux/iova.h b/include/linux/iova.h index dfa51ae49666..f1622ba17a27 100644 --- a/include/linux/iova.h +++ b/include/linux/iova.h @@ -26,12 +26,10 @@ struct iova_magazine; struct iova_cpu_rcache; #define IOVA_RANGE_CACHE_MAX_SIZE 6 /* log of max cached IOVA range size (in pages) */ -#define MAX_GLOBAL_MAGS 32 /* magazines per bin */ struct iova_rcache { spinlock_t lock; - unsigned long depot_size; - struct iova_magazine *depot[MAX_GLOBAL_MAGS]; + struct iova_magazine *depot; struct iova_cpu_rcache __percpu *cpu_rcaches; }; -- Gitee From 1deb56b774193c5e0f034baac6b0554fdafcd044 Mon Sep 17 00:00:00 2001 From: Zhang Zekun Date: Sun, 10 Dec 2023 14:10:35 +0800 Subject: [PATCH 6/6] iommu/iova: Manage the depot list size mainline inclusion from mainline-v6.7-rc1 commit 233045378dbbc0a7346127d19a54d4f91e0bd855 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8KS9A CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=233045378dbbc0a7346127d19a54d4f91e0bd855 ---------------------------------------- Automatically scaling the depot up to suit the peak capacity of a workload is all well and good, but it would be nice to have a way to scale it back down again if the workload changes. To that end, add backround reclaim that will gradually free surplus magazines if the depot size remains above a reasonable threshold for long enough. Reviewed-by: Jerry Snitselaar Signed-off-by: Robin Murphy Link: https://lore.kernel.org/r/03170665c56d89c6ce6081246b47f68d4e483308.1694535580.git.robin.murphy@arm.com Signed-off-by: Joerg Roedel Signed-off-by: Zhang Zekun (cherry picked from commit 4dfc6967f060bce5eb59790913cfd5ee8267876c) --- drivers/iommu/iova.c | 27 +++++++++++++++++++++++++++ include/linux/iova.h | 3 +++ 2 files changed, 30 insertions(+) diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c index 09ee0307f9f9..1f4a3fc56a6a 100644 --- a/drivers/iommu/iova.c +++ b/drivers/iommu/iova.c @@ -11,6 +11,7 @@ #include #include #include +#include /* The anchor node sits above the top of the usable address space */ #define IOVA_ANCHOR ~0UL @@ -793,6 +794,8 @@ split_and_remove_iova(struct iova_domain *iovad, struct iova *iova, */ #define IOVA_MAG_SIZE 127 +#define IOVA_DEPOT_DELAY msecs_to_jiffies(100) + struct iova_magazine { union { unsigned long size; @@ -886,6 +889,7 @@ static struct iova_magazine *iova_depot_pop(struct iova_rcache *rcache) rcache->depot = mag->next; mag->size = IOVA_MAG_SIZE; + rcache->depot_size--; return mag; } @@ -893,6 +897,25 @@ static void iova_depot_push(struct iova_rcache *rcache, struct iova_magazine *ma { mag->next = rcache->depot; rcache->depot = mag; + rcache->depot_size++; +} + +static void iova_depot_work_func(struct work_struct *work) +{ + struct iova_rcache *rcache = container_of(work, typeof(*rcache), work.work); + struct iova_magazine *mag = NULL; + unsigned long flags; + + spin_lock_irqsave(&rcache->lock, flags); + if (rcache->depot_size > num_online_cpus()) + mag = iova_depot_pop(rcache); + spin_unlock_irqrestore(&rcache->lock, flags); + + if (mag) { + iova_magazine_free_pfns(mag, rcache->iovad); + iova_magazine_free(mag); + schedule_delayed_work(&rcache->work, IOVA_DEPOT_DELAY); + } } static void init_iova_rcaches(struct iova_domain *iovad) @@ -905,6 +928,8 @@ static void init_iova_rcaches(struct iova_domain *iovad) for (i = 0; i < IOVA_RANGE_CACHE_MAX_SIZE; ++i) { rcache = &iovad->rcaches[i]; spin_lock_init(&rcache->lock); + rcache->iovad = iovad; + INIT_DELAYED_WORK(&rcache->work, iova_depot_work_func); rcache->cpu_rcaches = __alloc_percpu(sizeof(*cpu_rcache), cache_line_size()); if (WARN_ON(!rcache->cpu_rcaches)) continue; @@ -946,6 +971,7 @@ static bool __iova_rcache_insert(struct iova_domain *iovad, spin_lock(&rcache->lock); iova_depot_push(rcache, cpu_rcache->loaded); spin_unlock(&rcache->lock); + schedule_delayed_work(&rcache->work, IOVA_DEPOT_DELAY); cpu_rcache->loaded = new_mag; can_insert = true; @@ -1045,6 +1071,7 @@ static void free_iova_rcaches(struct iova_domain *iovad) iova_magazine_free(cpu_rcache->prev); } free_percpu(rcache->cpu_rcaches); + cancel_delayed_work_sync(&rcache->work); while (rcache->depot) iova_magazine_free(iova_depot_pop(rcache)); } diff --git a/include/linux/iova.h b/include/linux/iova.h index f1622ba17a27..25c447124638 100644 --- a/include/linux/iova.h +++ b/include/linux/iova.h @@ -29,8 +29,11 @@ struct iova_cpu_rcache; struct iova_rcache { spinlock_t lock; + unsigned int depot_size; struct iova_magazine *depot; struct iova_cpu_rcache __percpu *cpu_rcaches; + struct iova_domain *iovad; + struct delayed_work work; }; struct iova_domain; -- Gitee