Re: [PATCH V15 11/11] x86,cgroup/intel_rdt : Add a cgroup interface to manage Intel cache allocation

From: Marcelo Tosatti
Date: Wed Nov 18 2015 - 17:15:53 EST


On Thu, Oct 01, 2015 at 11:09:45PM -0700, Fenghua Yu wrote:
> Add a new cgroup 'intel_rdt' to manage cache allocation. Each cgroup
> directory is associated with a class of service id(closid). To map a
> task with closid during scheduling, this patch removes the closid field
> from task_struct and uses the already existing 'cgroups' field in
> task_struct.
>
> The cgroup has a file 'l3_cbm' which represents the L3 cache capacity
> bitmask(CBM). The CBM is global for the whole system currently. The
> capacity bitmask needs to have only contiguous bits set and number of
> bits that can be set is less than the max bits that can be set. The
> tasks belonging to a cgroup get to fill in the L3 cache represented by
> the capacity bitmask of the cgroup. For ex: if the max bits in the CBM
> is 10 and the cache size is 10MB, each bit represents 1MB of cache
> capacity.
>
> Root cgroup always has all the bits set in the l3_cbm. User can create
> more cgroups with mkdir syscall. By default the child cgroups inherit
> the capacity bitmask(CBM) from parent. User can change the CBM specified
> in hex for each cgroup. Each unique bitmask is associated with a class
> of service ID and an -ENOSPC is returned once we run out of
> closids.
>
> Signed-off-by: Vikas Shivappa <vikas.shivappa@xxxxxxxxxxxxxxx>
> Signed-off-by: Fenghua Yu <fenghua.yu@xxxxxxxxx>
> ---
> arch/x86/include/asm/intel_rdt.h | 37 +++++++-
> arch/x86/kernel/cpu/intel_rdt.c | 194 +++++++++++++++++++++++++++++++++++++--
> include/linux/cgroup_subsys.h | 4 +
> include/linux/sched.h | 3 -
> init/Kconfig | 4 +-
> 5 files changed, 229 insertions(+), 13 deletions(-)
>
> diff --git a/arch/x86/include/asm/intel_rdt.h b/arch/x86/include/asm/intel_rdt.h
> index afb6da3..fbe1e00 100644
> --- a/arch/x86/include/asm/intel_rdt.h
> +++ b/arch/x86/include/asm/intel_rdt.h
> @@ -3,6 +3,7 @@
>
> #ifdef CONFIG_INTEL_RDT
>
> +#include <linux/cgroup.h>
> #include <linux/jump_label.h>
>
> #define MAX_CBM_LENGTH 32
> @@ -12,20 +13,54 @@
> extern struct static_key rdt_enable_key;
> void __intel_rdt_sched_in(void *dummy);
>
> +struct intel_rdt {
> + struct cgroup_subsys_state css;
> + u32 closid;
> +};
> +
> struct clos_cbm_table {
> unsigned long l3_cbm;
> unsigned int clos_refcnt;
> };
>
> /*
> + * Return rdt group corresponding to this container.
> + */
> +static inline struct intel_rdt *css_rdt(struct cgroup_subsys_state *css)
> +{
> + return css ? container_of(css, struct intel_rdt, css) : NULL;
> +}
> +
> +static inline struct intel_rdt *parent_rdt(struct intel_rdt *ir)
> +{
> + return css_rdt(ir->css.parent);
> +}
> +
> +/*
> + * Return rdt group to which this task belongs.
> + */
> +static inline struct intel_rdt *task_rdt(struct task_struct *task)
> +{
> + return css_rdt(task_css(task, intel_rdt_cgrp_id));
> +}
> +
> +/*
> * intel_rdt_sched_in() - Writes the task's CLOSid to IA32_PQR_MSR
> *
> * Following considerations are made so that this has minimal impact
> * on scheduler hot path:
> * - This will stay as no-op unless we are running on an Intel SKU
> * which supports L3 cache allocation.
> + * - When support is present and enabled, does not do any
> + * IA32_PQR_MSR writes until the user starts really using the feature
> + * ie creates a rdt cgroup directory and assigns a cache_mask thats
> + * different from the root cgroup's cache_mask.
> * - Caches the per cpu CLOSid values and does the MSR write only
> - * when a task with a different CLOSid is scheduled in.

Why is this even allowed?

socket CBM bits:

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
[ | | | | | | | | | | | | | | ]

x x x x x x x
x x x x

x x x x x

cgroupA.bits = [ 0 - 6 ] cgroupB.bits = [ 10 - 14] (level 1)
cgroupA-A.bits = [ 0 - 4 ] (level 2)

Two ways to create a cgroup with bits [ 0 - 4] set:

1) Create a cgroup C in level 1 with a different name.
Useful to have same cgroup with two different names.

2) Create a cgroup A-B under cgroup-A with bits [0 - 4].

It just creates confusion, having two or more cgroups under
different levels of the hierarchy with the same bits set.
(can't see any organizational benefit).

Why not return -EINVAL ? Ah, cgroups are hierarchical, right.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/