[PATCH v2] cgroup/dmem: introduce a peak file

[PATCH v2] cgroup/dmem: introduce a peak file

[Thadeu Lima de Souza Cascardo]

Just like we have memory.peak, introduce a dmem.peak, which uses the
page_counter support for that.

For now, make it read-only.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@igalia.com>
---
Changes in v2:
- Make it read-only for now and adjust documentation accordingly.
- Link to v1: https://patch.msgid.link/20260506-dmem_peak-v1-0-8d803eb3449c@igalia.com
---
 Documentation/admin-guide/cgroup-v2.rst |  6 ++++++
 kernel/cgroup/dmem.c                    | 15 +++++++++++++++
 2 files changed, 21 insertions(+)

diff --git a/Documentation/admin-guide/cgroup-v2.rst b/Documentation/admin-guide/cgroup-v2.rst
index 6efd0095ed99..d103623b2be4 100644
--- a/Documentation/admin-guide/cgroup-v2.rst
+++ b/Documentation/admin-guide/cgroup-v2.rst
@@ -2808,6 +2808,12 @@ DMEM Interface Files
 	The semantics are the same as for the memory cgroup controller, and are
 	calculated in the same way.
 
+  dmem.peak
+	A read-only nested-keyed file that exists on non-root cgroups.
+
+	The max device memory usage recorded for the cgroup and its
+	descendants since the creation of the cgroup for each region.
+
   dmem.capacity
 	A read-only file that describes maximum region capacity.
 	It only exists on the root cgroup. Not all memory can be
diff --git a/kernel/cgroup/dmem.c b/kernel/cgroup/dmem.c
index 4753a67d0f0f..6430c7ce1e03 100644
--- a/kernel/cgroup/dmem.c
+++ b/kernel/cgroup/dmem.c
@@ -182,6 +182,11 @@ static u64 get_resource_current(struct dmem_cgroup_pool_state *pool)
 	return pool ? page_counter_read(&pool->cnt) : 0;
 }
 
+static u64 get_resource_peak(struct dmem_cgroup_pool_state *pool)
+{
+	return pool ? READ_ONCE(pool->cnt.watermark) : 0;
+}
+
 static void reset_all_resource_limits(struct dmem_cgroup_pool_state *rpool)
 {
 	set_resource_min(rpool, 0);
@@ -808,6 +813,11 @@ static int dmemcg_limit_show(struct seq_file *sf, void *v,
 	return 0;
 }
 
+static int dmem_cgroup_region_peak_show(struct seq_file *sf, void *v)
+{
+	return dmemcg_limit_show(sf, v, get_resource_peak);
+}
+
 static int dmem_cgroup_region_current_show(struct seq_file *sf, void *v)
 {
 	return dmemcg_limit_show(sf, v, get_resource_current);
@@ -856,6 +866,11 @@ static struct cftype files[] = {
 		.name = "current",
 		.seq_show = dmem_cgroup_region_current_show,
 	},
+	{
+		.name = "peak",
+		.seq_show = dmem_cgroup_region_peak_show,
+		.flags = CFTYPE_NOT_ON_ROOT,
+	},
 	{
 		.name = "min",
 		.write = dmem_cgroup_region_min_write,

---
base-commit: d3b0a7f21119f5a66cb76aa28fb8cc13206aaf7d
change-id: 20260409-dmem_peak-3abc1be95072

Best regards,
--  
Thadeu Lima de Souza Cascardo <cascardo@igalia.com>

Re: [PATCH v2] cgroup/dmem: introduce a peak file

[Tejun Heo]

Hello,

The patch looks fine to me, but please flesh out the motivation in the
commit description - what's the use case, why do we want this?

Thanks.

--
tejun

Claude review: cgroup/dmem: introduce a peak file

[Claude Code Review Bot]

Overall Series Review

Subject: cgroup/dmem: introduce a peak file
Author: Thadeu Lima de Souza Cascardo <cascardo@igalia.com>
Patches: 2
Reviewed: 2026-05-16T11:20:49.360654

---

This is a clean, well-scoped single patch that adds a `dmem.peak` cgroup file to the device memory (dmem) cgroup controller, mirroring the existing `memory.peak` pattern. The implementation correctly reuses the existing `dmemcg_limit_show` callback infrastructure and the `page_counter` watermark tracking that is already in place. The patch is minimal and follows established patterns in the codebase.

**Verdict: Looks good overall.** A couple of minor observations below but nothing blocking.

---
Generated by Claude Code Patch Reviewer

Claude review: cgroup/dmem: introduce a peak file

[Claude Code Review Bot]

Patch Review

**Approach:** The patch adds a `get_resource_peak()` getter that reads `pool->cnt.watermark`, a `dmem_cgroup_region_peak_show()` display function, and a cftype entry — all following the exact same pattern as `get_resource_current()` / `dmem_cgroup_region_current_show()`.

**Positive:**

1. The `READ_ONCE` usage in `get_resource_peak()` is correct and consistent with how `page_counter_charge()` in `mm/page_counter.c` uses `WRITE_ONCE` to update the watermark. This is the same deliberate-racy-but-acceptable pattern documented in the page counter code.

2. The null check (`pool ? ... : 0`) mirrors `get_resource_current()` exactly.

3. `CFTYPE_NOT_ON_ROOT` is the correct flag — this matches `memory.peak` in the memory controller and is consistent with how `dmem.min`, `dmem.low`, and `dmem.max` are configured.

**Observations:**

1. **`current` lacks `CFTYPE_NOT_ON_ROOT` but `peak` has it.** Looking at the existing tree, `dmem.current` does *not* set `CFTYPE_NOT_ON_ROOT`, so it appears on root. The new `dmem.peak` does set `CFTYPE_NOT_ON_ROOT`, so it won't appear on root. This is consistent with how `memory.current` (no flag) and `memory.peak` (has `CFTYPE_NOT_ON_ROOT`) work in `mm/memcontrol.c`, so the asymmetry is intentional and correct.

2. **No write/reset support.** The commit message and documentation say "read-only for now," which is fine for an initial version. The `memory.peak` controller has grown `peak_open`/`peak_release`/`memory_peak_write` for per-fd local watermark tracking and reset support. If that's planned as a follow-up, it might be worth mentioning in the commit message for future reference, but this is a nit — the v1→v2 changelog already documents the deliberate decision to start read-only.

3. **Documentation placement is correct.** The `dmem.peak` entry is inserted between `dmem.max/min/low` and `dmem.capacity`, which is a natural location grouping the per-cgroup files together before the root-only `dmem.capacity`.

4. **Watermark initialization.** The `page_counter_init()` function zeroes `usage` and sets `max` but does not explicitly initialize `watermark` or `local_watermark`. Since `page_counter` structs in `dmem_cgroup_pool_state` are allocated with `kzalloc` (which zero-initializes), the watermark starts at 0, which is correct for a "max seen so far" semantic. This is fine but worth noting for anyone reading the patch.

**No functional issues found.** The patch is a straightforward, correct extension of the existing dmem cgroup interface.

---
Generated by Claude Code Patch Reviewer