From: Shanker Donthineni <[email protected]>
The NUMA node information is visible to ITS driver but not being used
other than handling hardware errata. ITS/GICR hardware accesses to the
local NUMA node is usually quicker than the remote NUMA node. How slow
the remote NUMA accesses are depends on the implementation details.
This patch allocates memory for ITS management tables and command
queue from the corresponding NUMA node using the appropriate NUMA
aware functions. This change improves the performance of the ITS
tables read latency on systems where it has more than one ITS block,
and with the slower inter node accesses.
Apache Web server benchmarking using ab tool on a HiSilicon D06
board with multiple numa mem nodes shows Time per request and
Transfer rate improvements of ~3.6% with this patch.
Signed-off-by: Shanker Donthineni <[email protected]>
Signed-off-by: Hanjun Guo <[email protected]>
Signed-off-by: Shameer Kolothum <[email protected]>
Reviewed-by: Ganapatrao Kulkarni <[email protected]>
---
This is to revive the patch originally sent by Shanker[1] and
to back it up with a benchmark test. Any further testing of
this is most welcome.
v3-->v4
-Addressed comments on alloc_pages_node() and page_address() usage.
-Rebased on 5.0-rc1
-Added Ganapatrao's R-by.
v2-->v3
-Addressed comments to use page_address().
-Added Benchmark results to commit log.
-Removed T-by from Ganapatrao for now.
v1-->v2
-Edited commit text.
-Added Ganapatrao's tested-by.
Benchmark test details:
--------------------------------
Test Setup:
-D06 with dimm on node 0(Sock#0) and 3 (Sock#1).
-ITS belongs to numa node 0.
-Filesystem mounted on a PCIe NVMe based disk.
-Apache server installed on D06.
-Running ab benchmark test in concurrency mode from a remote m/c
connected to D06 via hns3(PCIe) n/w port.
"ab -k -c 750 -n 2000000 http://10.202.225.188/"
Test results are avg. of 15 runs.
For 4.20-rc1 Kernel,
----------------------------
Time per request(mean, concurrent) = 0.02753[ms]
Transfer Rate = 416501[Kbytes/sec]
For 4.20-rc1 + this patch,
----------------------------------
Time per request(mean, concurrent) = 0.02653[ms]
Transfer Rate = 431954[Kbytes/sec]
% improvement ~3.6%
vmstat shows around 170K-200K interrupts per second.
~# vmstat 1 -w
procs -----------------------memory-- - -system--
r b swpd free in
5 0 0 30166724 102794
9 0 0 30141828 171148
5 0 0 30150160 207185
13 0 0 30145924 175691
15 0 0 30140792 145250
13 0 0 30135556 201879
13 0 0 30134864 192391
10 0 0 30133632 168880
....
[1] https://patchwork.kernel.org/patch/9833339/
drivers/irqchip/irq-gic-v3-its.c | 26 ++++++++++++++++----------
1 file changed, 16 insertions(+), 10 deletions(-)
diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c
index db20e99..5df59ad 100644
--- a/drivers/irqchip/irq-gic-v3-its.c
+++ b/drivers/irqchip/irq-gic-v3-its.c
@@ -1737,6 +1737,7 @@ static int its_setup_baser(struct its_node *its, struct its_baser *baser,
u64 type = GITS_BASER_TYPE(val);
u64 baser_phys, tmp;
u32 alloc_pages;
+ struct page *page;
void *base;
retry_alloc_baser:
@@ -1749,10 +1750,11 @@ static int its_setup_baser(struct its_node *its, struct its_baser *baser,
order = get_order(GITS_BASER_PAGES_MAX * psz);
}
- base = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, order);
- if (!base)
+ page = alloc_pages_node(its->numa_node, GFP_KERNEL | __GFP_ZERO, order);
+ if (!page)
return -ENOMEM;
+ base = (void *)page_address(page);
baser_phys = virt_to_phys(base);
/* Check if the physical address of the memory is above 48bits */
@@ -2236,7 +2238,8 @@ static struct its_baser *its_get_baser(struct its_node *its, u32 type)
return NULL;
}
-static bool its_alloc_table_entry(struct its_baser *baser, u32 id)
+static bool its_alloc_table_entry(struct its_node *its,
+ struct its_baser *baser, u32 id)
{
struct page *page;
u32 esz, idx;
@@ -2256,7 +2259,8 @@ static bool its_alloc_table_entry(struct its_baser *baser, u32 id)
/* Allocate memory for 2nd level table */
if (!table[idx]) {
- page = alloc_pages(GFP_KERNEL | __GFP_ZERO, get_order(baser->psz));
+ page = alloc_pages_node(its->numa_node, GFP_KERNEL | __GFP_ZERO,
+ get_order(baser->psz));
if (!page)
return false;
@@ -2287,7 +2291,7 @@ static bool its_alloc_device_table(struct its_node *its, u32 dev_id)
if (!baser)
return (ilog2(dev_id) < its->device_ids);
- return its_alloc_table_entry(baser, dev_id);
+ return its_alloc_table_entry(its, baser, dev_id);
}
static bool its_alloc_vpe_table(u32 vpe_id)
@@ -2311,7 +2315,7 @@ static bool its_alloc_vpe_table(u32 vpe_id)
if (!baser)
return false;
- if (!its_alloc_table_entry(baser, vpe_id))
+ if (!its_alloc_table_entry(its, baser, vpe_id))
return false;
}
@@ -2345,7 +2349,7 @@ static struct its_device *its_create_device(struct its_node *its, u32 dev_id,
nr_ites = max(2, nvecs);
sz = nr_ites * its->ite_size;
sz = max(sz, ITS_ITT_ALIGN) + ITS_ITT_ALIGN - 1;
- itt = kzalloc(sz, GFP_KERNEL);
+ itt = kzalloc_node(sz, GFP_KERNEL, its->numa_node);
if (alloc_lpis) {
lpi_map = its_lpi_alloc(nvecs, &lpi_base, &nr_lpis);
if (lpi_map)
@@ -3486,6 +3490,7 @@ static int __init its_probe_one(struct resource *res,
void __iomem *its_base;
u32 val, ctlr;
u64 baser, tmp, typer;
+ struct page *page;
int err;
its_base = ioremap(res->start, resource_size(res));
@@ -3541,12 +3546,13 @@ static int __init its_probe_one(struct resource *res,
its->numa_node = numa_node;
- its->cmd_base = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
- get_order(ITS_CMD_QUEUE_SZ));
- if (!its->cmd_base) {
+ page = alloc_pages_node(its->numa_node, GFP_KERNEL | __GFP_ZERO,
+ get_order(ITS_CMD_QUEUE_SZ));
+ if (!page) {
err = -ENOMEM;
goto out_free_its;
}
+ its->cmd_base = (void *)page_address(page);
its->cmd_write = its->cmd_base;
its->fwnode_handle = handle;
its->get_msi_base = its_irq_get_msi_base;
--
2.7.4
Hi Marc,
A gentle reminder on this one...
Thanks,
Shameer
> -----Original Message-----
> From: Linuxarm [mailto:[email protected]] On Behalf Of Shameer
> Kolothum
> Sent: 14 January 2019 09:50
> To: [email protected]; [email protected]
> Cc: [email protected]; [email protected]; Linuxarm
> <[email protected]>; [email protected];
> [email protected]; [email protected]
> Subject: [PATCH v4] irqchip: gicv3-its: Use NUMA aware memory allocation for
> ITS tables
>
> From: Shanker Donthineni <[email protected]>
>
> The NUMA node information is visible to ITS driver but not being used
> other than handling hardware errata. ITS/GICR hardware accesses to the
> local NUMA node is usually quicker than the remote NUMA node. How slow
> the remote NUMA accesses are depends on the implementation details.
>
> This patch allocates memory for ITS management tables and command
> queue from the corresponding NUMA node using the appropriate NUMA
> aware functions. This change improves the performance of the ITS
> tables read latency on systems where it has more than one ITS block,
> and with the slower inter node accesses.
>
> Apache Web server benchmarking using ab tool on a HiSilicon D06
> board with multiple numa mem nodes shows Time per request and
> Transfer rate improvements of ~3.6% with this patch.
>
> Signed-off-by: Shanker Donthineni <[email protected]>
> Signed-off-by: Hanjun Guo <[email protected]>
> Signed-off-by: Shameer Kolothum <[email protected]>
> Reviewed-by: Ganapatrao Kulkarni <[email protected]>
> ---
>
> This is to revive the patch originally sent by Shanker[1] and
> to back it up with a benchmark test. Any further testing of
> this is most welcome.
>
> v3-->v4
> -Addressed comments on alloc_pages_node() and page_address() usage.
> -Rebased on 5.0-rc1
> -Added Ganapatrao's R-by.
>
> v2-->v3
> -Addressed comments to use page_address().
> -Added Benchmark results to commit log.
> -Removed T-by from Ganapatrao for now.
>
> v1-->v2
> -Edited commit text.
> -Added Ganapatrao's tested-by.
>
> Benchmark test details:
> --------------------------------
> Test Setup:
> -D06 with dimm on node 0(Sock#0) and 3 (Sock#1).
> -ITS belongs to numa node 0.
> -Filesystem mounted on a PCIe NVMe based disk.
> -Apache server installed on D06.
> -Running ab benchmark test in concurrency mode from a remote m/c
> connected to D06 via hns3(PCIe) n/w port.
> "ab -k -c 750 -n 2000000 http://10.202.225.188/"
>
> Test results are avg. of 15 runs.
>
> For 4.20-rc1 Kernel,
> ----------------------------
> Time per request(mean, concurrent) = 0.02753[ms]
> Transfer Rate = 416501[Kbytes/sec]
>
> For 4.20-rc1 + this patch,
> ----------------------------------
> Time per request(mean, concurrent) = 0.02653[ms]
> Transfer Rate = 431954[Kbytes/sec]
>
> % improvement ~3.6%
>
> vmstat shows around 170K-200K interrupts per second.
>
> ~# vmstat 1 -w
> procs -----------------------memory-- - -system--
> r b swpd free in
> 5 0 0 30166724 102794
> 9 0 0 30141828 171148
> 5 0 0 30150160 207185
> 13 0 0 30145924 175691
> 15 0 0 30140792 145250
> 13 0 0 30135556 201879
> 13 0 0 30134864 192391
> 10 0 0 30133632 168880
> ....
>
> [1] https://patchwork.kernel.org/patch/9833339/
>
> drivers/irqchip/irq-gic-v3-its.c | 26 ++++++++++++++++----------
> 1 file changed, 16 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c
> index db20e99..5df59ad 100644
> --- a/drivers/irqchip/irq-gic-v3-its.c
> +++ b/drivers/irqchip/irq-gic-v3-its.c
> @@ -1737,6 +1737,7 @@ static int its_setup_baser(struct its_node *its,
> struct its_baser *baser,
> u64 type = GITS_BASER_TYPE(val);
> u64 baser_phys, tmp;
> u32 alloc_pages;
> + struct page *page;
> void *base;
>
> retry_alloc_baser:
> @@ -1749,10 +1750,11 @@ static int its_setup_baser(struct its_node *its,
> struct its_baser *baser,
> order = get_order(GITS_BASER_PAGES_MAX * psz);
> }
>
> - base = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, order);
> - if (!base)
> + page = alloc_pages_node(its->numa_node, GFP_KERNEL | __GFP_ZERO,
> order);
> + if (!page)
> return -ENOMEM;
>
> + base = (void *)page_address(page);
> baser_phys = virt_to_phys(base);
>
> /* Check if the physical address of the memory is above 48bits */
> @@ -2236,7 +2238,8 @@ static struct its_baser *its_get_baser(struct
> its_node *its, u32 type)
> return NULL;
> }
>
> -static bool its_alloc_table_entry(struct its_baser *baser, u32 id)
> +static bool its_alloc_table_entry(struct its_node *its,
> + struct its_baser *baser, u32 id)
> {
> struct page *page;
> u32 esz, idx;
> @@ -2256,7 +2259,8 @@ static bool its_alloc_table_entry(struct its_baser
> *baser, u32 id)
>
> /* Allocate memory for 2nd level table */
> if (!table[idx]) {
> - page = alloc_pages(GFP_KERNEL | __GFP_ZERO,
> get_order(baser->psz));
> + page = alloc_pages_node(its->numa_node, GFP_KERNEL |
> __GFP_ZERO,
> + get_order(baser->psz));
> if (!page)
> return false;
>
> @@ -2287,7 +2291,7 @@ static bool its_alloc_device_table(struct its_node
> *its, u32 dev_id)
> if (!baser)
> return (ilog2(dev_id) < its->device_ids);
>
> - return its_alloc_table_entry(baser, dev_id);
> + return its_alloc_table_entry(its, baser, dev_id);
> }
>
> static bool its_alloc_vpe_table(u32 vpe_id)
> @@ -2311,7 +2315,7 @@ static bool its_alloc_vpe_table(u32 vpe_id)
> if (!baser)
> return false;
>
> - if (!its_alloc_table_entry(baser, vpe_id))
> + if (!its_alloc_table_entry(its, baser, vpe_id))
> return false;
> }
>
> @@ -2345,7 +2349,7 @@ static struct its_device *its_create_device(struct
> its_node *its, u32 dev_id,
> nr_ites = max(2, nvecs);
> sz = nr_ites * its->ite_size;
> sz = max(sz, ITS_ITT_ALIGN) + ITS_ITT_ALIGN - 1;
> - itt = kzalloc(sz, GFP_KERNEL);
> + itt = kzalloc_node(sz, GFP_KERNEL, its->numa_node);
> if (alloc_lpis) {
> lpi_map = its_lpi_alloc(nvecs, &lpi_base, &nr_lpis);
> if (lpi_map)
> @@ -3486,6 +3490,7 @@ static int __init its_probe_one(struct resource *res,
> void __iomem *its_base;
> u32 val, ctlr;
> u64 baser, tmp, typer;
> + struct page *page;
> int err;
>
> its_base = ioremap(res->start, resource_size(res));
> @@ -3541,12 +3546,13 @@ static int __init its_probe_one(struct resource
> *res,
>
> its->numa_node = numa_node;
>
> - its->cmd_base = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
> - get_order(ITS_CMD_QUEUE_SZ));
> - if (!its->cmd_base) {
> + page = alloc_pages_node(its->numa_node, GFP_KERNEL | __GFP_ZERO,
> + get_order(ITS_CMD_QUEUE_SZ));
> + if (!page) {
> err = -ENOMEM;
> goto out_free_its;
> }
> + its->cmd_base = (void *)page_address(page);
> its->cmd_write = its->cmd_base;
> its->fwnode_handle = handle;
> its->get_msi_base = its_irq_get_msi_base;
> --
> 2.7.4
>
>
> _______________________________________________
> Linuxarm mailing list
> [email protected]
> http://hulk.huawei.com/mailman/listinfo/linuxarm
On Tue, 19 Feb 2019 11:05:38 +0000,
Shameerali Kolothum Thodi <[email protected]> wrote:
>
> Hi Marc,
>
> A gentle reminder on this one...
Thanks for that. Now applied to irqchip-next.
M.
--
Jazz is not dead, it just smell funny.