Re: [PATCH v3 3/4] mm: shmem: override mTHP shmem default with a kernel parameter

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi David,

On 31/10/24 09:57, David Hildenbrand wrote:
On 31.10.24 13:51, Maíra Canal wrote:
Hi David,

On 31/10/24 09:37, David Hildenbrand wrote:
On 30.10.24 13:58, Maíra Canal wrote:
Add the ``thp_shmem=`` kernel command line to allow specifying the
default policy of each supported shmem hugepage size. The kernel
parameter
accepts the following format:

thp_shmem=<size>[KMG],<size>[KMG]:<policy>;<size>[KMG]-
<size>[KMG]:<policy>

For example,

thp_shmem=16K-64K:always;128K,512K:inherit;256K:advise;1M-2M:never;4M-8M:within_size

By configuring the default policy of several shmem hugepages, the user
can take advantage of mTHP before it's been configured through sysfs.

Signed-off-by: Maíra Canal <mcanal@xxxxxxxxxx>
---
   .../admin-guide/kernel-parameters.txt         |  10 ++
   Documentation/admin-guide/mm/transhuge.rst    |  17 +++
   mm/shmem.c                                    | 109 +++++++++++++ ++++-
   3 files changed, 135 insertions(+), 1 deletion(-)


[...]

diff --git a/mm/shmem.c b/mm/shmem.c
index dfcc88ec6e34..c2299fa0b345 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -136,6 +136,7 @@ static unsigned long huge_shmem_orders_always
__read_mostly;
   static unsigned long huge_shmem_orders_madvise __read_mostly;
   static unsigned long huge_shmem_orders_inherit __read_mostly;
   static unsigned long huge_shmem_orders_within_size __read_mostly;
+static bool shmem_orders_configured __initdata;
   #endif
   #ifdef CONFIG_TMPFS
@@ -5027,7 +5028,8 @@ void __init shmem_init(void)
        * Default to setting PMD-sized THP to inherit the global
setting and
        * disable all other multi-size THPs.
        */
-    huge_shmem_orders_inherit = BIT(HPAGE_PMD_ORDER);
+    if (!shmem_orders_configured)
+        huge_shmem_orders_inherit = BIT(HPAGE_PMD_ORDER);
   #endif
       return;
@@ -5180,6 +5182,26 @@ struct kobj_attribute thpsize_shmem_enabled_attr =
   #if defined(CONFIG_TRANSPARENT_HUGEPAGE)
+static inline int get_order_from_str(const char *size_str)
+{
+    unsigned long size;
+    char *endptr;
+    int order;
+
+    size = memparse(size_str, &endptr);
+
+    if (!is_power_of_2(size))
+        goto err;
+    order = get_order(size);
+    if (BIT(order) & ~THP_ORDERS_ALL_FILE_DEFAULT)
+        goto err;
+
+    return order;
+err:
+    pr_err("invalid size %s in thp_shmem boot parameter\n", size_str);
+    return -EINVAL;
+}

Hm, mostly copy and paste. You could reuse existing get_order_from_str()
simply by passing in the supported orders and moving error reporting to
the caller.


Can I use functions from mm/huge_memory.c here?

Yes, that's the idea.


Unfortunately, it isn't possible without adding the function to a
header.

I deleted `get_order_from_str()` from mm/shmem.c just to test it:

mm/shmem.c:5230:13: error: call to undeclared function 'get_order_from_str'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration] 5230 | start = get_order_from_str(start_size);
      |                                         ^
mm/shmem.c:5233:19: error: call to undeclared function 'get_order_from_str'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration] 5233 | start = end = get_order_from_str(subtoken);
      |                                               ^
2 errors generated.
make[3]: *** [scripts/Makefile.build:229: mm/shmem.o] Error 1
make[2]: *** [scripts/Makefile.build:478: mm] Error 2
make[2]: *** Waiting for unfinished jobs....

Please, check the discussion I had with Barry about the pros and cons of
copy and paste and code refactor [1][2]. We ended up deciding to
duplicate the code.

[1] https://lore.kernel.org/linux-mm/CAGsJ_4wJ2xoNRLMNkmLL3x1t2YiqQJ1=saXa+HNxRUbSNivRCw@xxxxxxxxxxxxxx/ [2] https://lore.kernel.org/linux-mm/CAGsJ_4yp89one_hoB_87poU2wrpJh_0NVicRKW538eE6yo1kZw@xxxxxxxxxxxxxx/

Best Regards,
- Maíra





[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux