On Fri, Aug 9, 2024 at 4:57 PM Ryan Roberts <ryan.roberts@xxxxxxx> wrote:
>
> On 09/08/2024 09:31, Barry Song wrote:
> > On Fri, Aug 9, 2024 at 3:58 PM Ryan Roberts <ryan.roberts@xxxxxxx> wrote:
> >>
> >> On 08/08/2024 22:17, Barry Song wrote:
> >>> On Thu, Aug 8, 2024 at 10:17 PM Ryan Roberts <ryan.roberts@xxxxxxx> wrote:
> >>>>
> >>>> Add thp_anon= cmdline parameter to allow specifying the default
> >>>> enablement of each supported anon THP size. The parameter accepts the
> >>>> following format and can be provided multiple times to configure each
> >>>> size:
> >>>>
> >>>> thp_anon=<size>[KMG]:<value>
> >>>>
> >>>> See Documentation/admin-guide/mm/transhuge.rst for more details.
> >>>>
> >>>> Configuring the defaults at boot time is useful to allow early user
> >>>> space to take advantage of mTHP before its been configured through
> >>>> sysfs.
> >>>>
> >>>> Signed-off-by: Ryan Roberts <ryan.roberts@xxxxxxx>
> >>>> ---
> >>>>
> >>>> Hi All,
> >>>>
> >>>> I've split this off from my RFC at [1] because Barry highlighted that he would
> >>>> benefit from it immediately [2]. There are no changes vs the version in that
> >>>> series.
> >>>>
> >>>> It applies against today's mm-unstable (275d686abcb59). (although I had to fix a
> >>>> minor build bug in stackdepot.c due to MIN() not being defined in this tree).
> >>>>
> >>>> Thanks,
> >>>> Ryan
> >>>>
> >>>>
> >>>> .../admin-guide/kernel-parameters.txt | 8 +++
> >>>> Documentation/admin-guide/mm/transhuge.rst | 26 +++++++--
> >>>> mm/huge_memory.c | 55 ++++++++++++++++++-
> >>>> 3 files changed, 82 insertions(+), 7 deletions(-)
> >>>>
> >>>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> >>>> index bcdee8984e1f0..5c79b58c108ec 100644
> >>>> --- a/Documentation/admin-guide/kernel-parameters.txt
> >>>> +++ b/Documentation/admin-guide/kernel-parameters.txt
> >>>> @@ -6631,6 +6631,14 @@
> >>>> <deci-seconds>: poll all this frequency
> >>>> 0: no polling (default)
> >>>>
> >>>> + thp_anon= [KNL]
> >>>> + Format: <size>[KMG]:always|madvise|never|inherit
> >>>> + Can be used to control the default behavior of the
> >>>> + system with respect to anonymous transparent hugepages.
> >>>> + Can be used multiple times for multiple anon THP sizes.
> >>>> + See Documentation/admin-guide/mm/transhuge.rst for more
> >>>> + details.
> >>>> +
> >>>> threadirqs [KNL,EARLY]
> >>>> Force threading of all interrupt handlers except those
> >>>> marked explicitly IRQF_NO_THREAD.
> >>>> diff --git a/Documentation/admin-guide/mm/transhuge.rst b/Documentation/admin-guide/mm/transhuge.rst
> >>>> index 24eec1c03ad88..f63b0717366c6 100644
> >>>> --- a/Documentation/admin-guide/mm/transhuge.rst
> >>>> +++ b/Documentation/admin-guide/mm/transhuge.rst
> >>>> @@ -284,13 +284,27 @@ that THP is shared. Exceeding the number would block the collapse::
> >>>>
> >>>> A higher value may increase memory footprint for some workloads.
> >>>>
> >>>> -Boot parameter
> >>>> -==============
> >>>> +Boot parameters
> >>>> +===============
> >>>>
> >>>> -You can change the sysfs boot time defaults of Transparent Hugepage
> >>>> -Support by passing the parameter ``transparent_hugepage=always`` or
> >>>> -``transparent_hugepage=madvise`` or ``transparent_hugepage=never``
> >>>> -to the kernel command line.
> >>>> +You can change the sysfs boot time default for the top-level "enabled"
> >>>> +control by passing the parameter ``transparent_hugepage=always`` or
> >>>> +``transparent_hugepage=madvise`` or ``transparent_hugepage=never`` to the
> >>>> +kernel command line.
> >>>> +
> >>>> +Alternatively, each supported anonymous THP size can be controlled by
> >>>> +passing ``thp_anon=<size>[KMG]:<state>``, where ``<size>`` is the THP size
> >>>> +and ``<state>`` is one of ``always``, ``madvise``, ``never`` or
> >>>> +``inherit``.
> >>>> +
> >>>> +For example, the following will set 64K THP to ``always``::
> >>>> +
> >>>> + thp_anon=64K:always
> >>>> +
> >>>> +``thp_anon=`` may be specified multiple times to configure all THP sizes as
> >>>> +required. If ``thp_anon=`` is specified at least once, any anon THP sizes
> >>>> +not explicitly configured on the command line are implicitly set to
> >>>> +``never``.
> >>>>
> >>>> Hugepages in tmpfs/shmem
> >>>> ========================
> >>>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> >>>> index 0c3075ee00012..c2c0da1eb94e6 100644
> >>>> --- a/mm/huge_memory.c
> >>>> +++ b/mm/huge_memory.c
> >>>> @@ -82,6 +82,7 @@ unsigned long huge_zero_pfn __read_mostly = ~0UL;
> >>>> unsigned long huge_anon_orders_always __read_mostly;
> >>>> unsigned long huge_anon_orders_madvise __read_mostly;
> >>>> unsigned long huge_anon_orders_inherit __read_mostly;
> >>>> +static bool anon_orders_configured;
> >>>>
> >>>> unsigned long __thp_vma_allowable_orders(struct vm_area_struct *vma,
> >>>> unsigned long vm_flags,
> >>>> @@ -672,7 +673,10 @@ static int __init hugepage_init_sysfs(struct kobject **hugepage_kobj)
> >>>> * disable all other sizes. powerpc's PMD_ORDER isn't a compile-time
> >>>> * constant so we have to do this here.
> >>>> */
> >>>> - huge_anon_orders_inherit = BIT(PMD_ORDER);
> >>>> + if (!anon_orders_configured) {
> >>>> + huge_anon_orders_inherit = BIT(PMD_ORDER);
> >>>> + anon_orders_configured = true;
> >>>> + }
> >>>
> >>> If a user configures 64KB and doesn't adjust anything for PMD_ORDER,
> >>> then PMD_ORDER will be set to "never", correct? This seems to change
> >>> the default behavior of PMD_ORDER. Could we instead achieve this by
> >>> checking if PMD_ORDER has been explicitly configured?
> >>
> >> Yes, that's how it's implemented in this patch, and the accompanying docs also
> >> state:
> >>
> >> If ``thp_anon=`` is specified at least once, any anon THP sizes
> >> not explicitly configured on the command line are implicitly set to
> >> ``never``.
> >>
> >> My initial approach did exactly as you suggest. But in the original series, I
> >> also had a similar patch to configure file thp with "thp_file=". And for file,
> >> all of the orders default to `always`. So if taking the same approach with that
> >> control, the user would have to explicitly opt-out of all supported orders
> >> rather than just opt-in to the orders they want. And I thought that could get
> >> tricky in future if support is added for more orders. I felt that was
> >> potentially very confusing so decided it was clearer to have the above rule and
> >> make both controls consistent.
> >>
> >> What do you think?
> >
> > If this is the intention, once the user sets the command line, they should
> > realize that the default settings have been overridden. I am perfectly fine
> > with this strategy.
>
> OK. I can see it from both sides to be honest. Let's see if anyone else has
> issue with this approach.
>
> >
> > with the below cmdline:
> > thp_anon=64K:always thp_anon=8K:inherit thp_anon=32K:madvise
> > thp_anon=1M:inherit thp_anon=2M:always
> >
> > I am getting:
> > / # cat /sys/kernel/mm/transparent_hugepage/hugepages-64kB/enabled
> > [always] inherit madvise never
> > / # cat /sys/kernel/mm/transparent_hugepage/hugepages-32kB/enabled
> > always inherit [madvise] never
> > / # cat /sys/kernel/mm/transparent_hugepage/hugepages-1024kB/enabled
> > always [inherit] madvise never
> > / # cat /sys/kernel/mm/transparent_hugepage/hugepages-2048kB/enabled
> > [always] inherit madvise never
>
> And you should also be seeing a warning in the boot log that thp_anon=8K:inherit
> is unrecognised (since anon doesn't support order-1).
Indeed.
I am even seeing "Unknown kernel command line parameters
"thp_anon=8K:inherit", will be passed to user space."
[ 0.000000] huge_memory: thp_anon=8K:inherit: cannot parse, ignored
[ 0.000000] Unknown kernel command line parameters
"thp_anon=8K:inherit", will be passed to user space.
[ 0.000000] random: crng init done
[ 0.000000] Dentry cache hash table entries: 65536 (order: 7,
524288 bytes, linear)
[ 0.000000] Inode-cache hash table entries: 32768 (order: 6, 262144
bytes, linear)
[ 0.000000] Fallback order for Node 0: 0
[ 0.000000] Built 1 zonelists, mobility grouping on. Total pages: 131072
[ 0.000000] Policy zone: DMA
>
> >
> > Thus,
> >
> > Tested-by: Barry Song <baohua@xxxxxxxxxx>
>
> Thanks!
>
> >
> >>
> >>
> >>>
> >>>>
> >>>> *hugepage_kobj = kobject_create_and_add("transparent_hugepage", mm_kobj);
> >>>> if (unlikely(!*hugepage_kobj)) {
> >>>> @@ -857,6 +861,55 @@ static int __init setup_transparent_hugepage(char *str)
> >>>> }
> >>>> __setup("transparent_hugepage=", setup_transparent_hugepage);
> >>>>
> >>>> +static int __init setup_thp_anon(char *str)
> >>>> +{
> >>>> + unsigned long size;
> >>>> + char *state;
> >>>> + int order;
> >>>> + int ret = 0;
> >>>> +
> >>>> + if (!str)
> >>>> + goto out;
> >>>> +
> >>>> + size = (unsigned long)memparse(str, &state);
> >>>> + order = ilog2(size >> PAGE_SHIFT);
> >>>> + if (*state != ':' || !is_power_of_2(size) || size <= PAGE_SIZE ||
> >>>> + !(BIT(order) & THP_ORDERS_ALL_ANON))
> >>>> + goto out;
> >>>> +
> >>>> + state++;
> >>>> +
> >>>> + if (!strcmp(state, "always")) {
> >>>> + clear_bit(order, &huge_anon_orders_inherit);
> >>>> + clear_bit(order, &huge_anon_orders_madvise);
> >>>> + set_bit(order, &huge_anon_orders_always);
> >>>> + ret = 1;
> >>>> + } else if (!strcmp(state, "inherit")) {
> >>>> + clear_bit(order, &huge_anon_orders_always);
> >>>> + clear_bit(order, &huge_anon_orders_madvise);
> >>>> + set_bit(order, &huge_anon_orders_inherit);
> >>>> + ret = 1;
> >>>> + } else if (!strcmp(state, "madvise")) {
> >>>> + clear_bit(order, &huge_anon_orders_always);
> >>>> + clear_bit(order, &huge_anon_orders_inherit);
> >>>> + set_bit(order, &huge_anon_orders_madvise);
> >>>> + ret = 1;
> >>>> + } else if (!strcmp(state, "never")) {
> >>>> + clear_bit(order, &huge_anon_orders_always);
> >>>> + clear_bit(order, &huge_anon_orders_inherit);
> >>>> + clear_bit(order, &huge_anon_orders_madvise);
> >>>> + ret = 1;
> >>>> + }
> >>>> +
> >>>> + if (ret)
> >>>> + anon_orders_configured = true;
> >>>
> >>> I mean:
> >>>
> >>> if (ret && order == PMD_ORDER)
> >>> anon_pmd_order_configured = true;
> >>>
> >>>> +out:
> >>>> + if (!ret)
> >>>> + pr_warn("thp_anon=%s: cannot parse, ignored\n", str);
> >>>> + return ret;
> >>>> +}
> >>>> +__setup("thp_anon=", setup_thp_anon);
> >>>> +
> >>>> pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma)
> >>>> {
> >>>> if (likely(vma->vm_flags & VM_WRITE))
> >>>> --
> >>>> 2.43.0
> >>>>
> >>>
> >>> Thanks
> >>> Barry
> >>
>
>
