From f1ab2831e2a4312046bca79256b2efc41d373eaf Mon Sep 17 00:00:00 2001 From: Tang Yizhou Date: Tue, 4 Mar 2025 19:03:16 +0800 Subject: writeback: let trace_balance_dirty_pages() take struct dtc as parameter Patch series "Fix calculations in trace_balance_dirty_pages() for cgwb", v2. In my experiment, I found that the output of trace_balance_dirty_pages() in the cgroup writeback scenario was strange because trace_balance_dirty_pages() always uses global_wb_domain.dirty_limit for related calculations instead of the dirty_limit of the corresponding memcg's wb_domain. The basic idea of the fix is to store the hard dirty limit value computed in wb_position_ratio() into struct dirty_throttle_control and use it for calculations in trace_balance_dirty_pages(). This patch (of 3): Currently, trace_balance_dirty_pages() already has 12 parameters. In the patch #3, I initially attempted to introduce an additional parameter. However, in include/linux/trace_events.h, bpf_trace_run12() only supports up to 12 parameters and bpf_trace_run13() does not exist. To reduce the number of parameters in trace_balance_dirty_pages(), we can make it accept a pointer to struct dirty_throttle_control as a parameter. To achieve this, we need to move the definition of struct dirty_throttle_control from mm/page-writeback.c to include/linux/writeback.h. Link: https://lkml.kernel.org/r/20250304110318.159567-1-yizhou.tang@shopee.com Link: https://lkml.kernel.org/r/20250304110318.159567-2-yizhou.tang@shopee.com Signed-off-by: Tang Yizhou Cc: Alexei Starovoitov Cc: Christian Brauner Cc: Steven Rostedt Cc: Jan Kara Cc: "Masami Hiramatsu (Google)" Cc: Matthew Wilcow (Oracle) Cc: Tang Yizhou Cc: Tejun Heo Signed-off-by: Andrew Morton --- include/trace/events/writeback.h | 16 ++++++---------- 1 file changed, 6 insertions(+), 10 deletions(-) (limited to 'include/trace') diff --git a/include/trace/events/writeback.h b/include/trace/events/writeback.h index a261e86e61fa..3213b9023794 100644 --- a/include/trace/events/writeback.h +++ b/include/trace/events/writeback.h @@ -629,11 +629,7 @@ TRACE_EVENT(bdi_dirty_ratelimit, TRACE_EVENT(balance_dirty_pages, TP_PROTO(struct bdi_writeback *wb, - unsigned long thresh, - unsigned long bg_thresh, - unsigned long dirty, - unsigned long bdi_thresh, - unsigned long bdi_dirty, + struct dirty_throttle_control *dtc, unsigned long dirty_ratelimit, unsigned long task_ratelimit, unsigned long dirtied, @@ -641,7 +637,7 @@ TRACE_EVENT(balance_dirty_pages, long pause, unsigned long start_time), - TP_ARGS(wb, thresh, bg_thresh, dirty, bdi_thresh, bdi_dirty, + TP_ARGS(wb, dtc, dirty_ratelimit, task_ratelimit, dirtied, period, pause, start_time), @@ -664,16 +660,16 @@ TRACE_EVENT(balance_dirty_pages, ), TP_fast_assign( - unsigned long freerun = (thresh + bg_thresh) / 2; + unsigned long freerun = (dtc->thresh + dtc->bg_thresh) / 2; strscpy_pad(__entry->bdi, bdi_dev_name(wb->bdi), 32); __entry->limit = global_wb_domain.dirty_limit; __entry->setpoint = (global_wb_domain.dirty_limit + freerun) / 2; - __entry->dirty = dirty; + __entry->dirty = dtc->dirty; __entry->bdi_setpoint = __entry->setpoint * - bdi_thresh / (thresh + 1); - __entry->bdi_dirty = bdi_dirty; + dtc->wb_thresh / (dtc->thresh + 1); + __entry->bdi_dirty = dtc->wb_dirty; __entry->dirty_ratelimit = KBps(dirty_ratelimit); __entry->task_ratelimit = KBps(task_ratelimit); __entry->dirtied = dirtied; -- cgit v1.2.3 From 28c24ef9e04f95672b72b1297eff7dae91cceea8 Mon Sep 17 00:00:00 2001 From: Tang Yizhou Date: Tue, 4 Mar 2025 19:03:17 +0800 Subject: writeback: rename variables in trace_balance_dirty_pages() Rename bdi_setpoint and bdi_dirty in the tracepoint to wb_setpoint and wb_dirty, respectively. These changes were omitted by Tejun in the cgroup writeback patchset. Link: https://lkml.kernel.org/r/20250304110318.159567-3-yizhou.tang@shopee.com Signed-off-by: Tang Yizhou Cc: Alexei Starovoitov Cc: Christian Brauner Cc: Jan Kara Cc: "Masami Hiramatsu (Google)" Cc: Matthew Wilcow (Oracle) Cc: Steven Rostedt Cc: Tejun Heo Signed-off-by: Andrew Morton --- include/trace/events/writeback.h | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'include/trace') diff --git a/include/trace/events/writeback.h b/include/trace/events/writeback.h index 3213b9023794..3046ca6b08ea 100644 --- a/include/trace/events/writeback.h +++ b/include/trace/events/writeback.h @@ -646,8 +646,8 @@ TRACE_EVENT(balance_dirty_pages, __field(unsigned long, limit) __field(unsigned long, setpoint) __field(unsigned long, dirty) - __field(unsigned long, bdi_setpoint) - __field(unsigned long, bdi_dirty) + __field(unsigned long, wb_setpoint) + __field(unsigned long, wb_dirty) __field(unsigned long, dirty_ratelimit) __field(unsigned long, task_ratelimit) __field(unsigned int, dirtied) @@ -667,9 +667,9 @@ TRACE_EVENT(balance_dirty_pages, __entry->setpoint = (global_wb_domain.dirty_limit + freerun) / 2; __entry->dirty = dtc->dirty; - __entry->bdi_setpoint = __entry->setpoint * + __entry->wb_setpoint = __entry->setpoint * dtc->wb_thresh / (dtc->thresh + 1); - __entry->bdi_dirty = dtc->wb_dirty; + __entry->wb_dirty = dtc->wb_dirty; __entry->dirty_ratelimit = KBps(dirty_ratelimit); __entry->task_ratelimit = KBps(task_ratelimit); __entry->dirtied = dirtied; @@ -685,7 +685,7 @@ TRACE_EVENT(balance_dirty_pages, TP_printk("bdi %s: " "limit=%lu setpoint=%lu dirty=%lu " - "bdi_setpoint=%lu bdi_dirty=%lu " + "wb_setpoint=%lu wb_dirty=%lu " "dirty_ratelimit=%lu task_ratelimit=%lu " "dirtied=%u dirtied_pause=%u " "paused=%lu pause=%ld period=%lu think=%ld cgroup_ino=%lu", @@ -693,8 +693,8 @@ TRACE_EVENT(balance_dirty_pages, __entry->limit, __entry->setpoint, __entry->dirty, - __entry->bdi_setpoint, - __entry->bdi_dirty, + __entry->wb_setpoint, + __entry->wb_dirty, __entry->dirty_ratelimit, __entry->task_ratelimit, __entry->dirtied, -- cgit v1.2.3 From 6cc4c3aa714bc58ec5d20f3054ca5f23534984d1 Mon Sep 17 00:00:00 2001 From: Tang Yizhou Date: Tue, 4 Mar 2025 19:03:18 +0800 Subject: writeback: fix calculations in trace_balance_dirty_pages() for cgwb In the commit dcc25ae76eb7 ("writeback: move global_dirty_limit into wb_domain") of the cgroup writeback backpressure propagation patchset, Tejun made some adaptations to trace_balance_dirty_pages() for cgroup writeback. However, this adaptation was incomplete and Tejun missed further adaptation in the subsequent patches. In the cgroup writeback scenario, if sdtc in balance_dirty_pages() is assigned to mdtc, then upon entering trace_balance_dirty_pages(), __entry->limit should be assigned based on the dirty_limit of the corresponding memcg's wb_domain, rather than global_wb_domain. To address this issue and simplify the implementation, introduce a 'limit' field in struct dirty_throttle_control to store the hard_limit value computed in wb_position_ratio() by calling hard_dirty_limit(). This field will then be used in trace_balance_dirty_pages() to assign the value to __entry->limit. Link: https://lkml.kernel.org/r/20250304110318.159567-4-yizhou.tang@shopee.com Fixes: dcc25ae76eb7 ("writeback: move global_dirty_limit into wb_domain") Signed-off-by: Tang Yizhou Acked-by: Tejun Heo Cc: Alexei Starovoitov Cc: Christian Brauner Cc: Jan Kara Cc: "Masami Hiramatsu (Google)" Cc: Matthew Wilcow (Oracle) Cc: Steven Rostedt Signed-off-by: Andrew Morton --- include/linux/writeback.h | 1 + include/trace/events/writeback.h | 5 ++--- mm/page-writeback.c | 2 +- 3 files changed, 4 insertions(+), 4 deletions(-) (limited to 'include/trace') diff --git a/include/linux/writeback.h b/include/linux/writeback.h index 32095928365c..58bda3347914 100644 --- a/include/linux/writeback.h +++ b/include/linux/writeback.h @@ -326,6 +326,7 @@ struct dirty_throttle_control { unsigned long dirty; /* file_dirty + write + nfs */ unsigned long thresh; /* dirty threshold */ unsigned long bg_thresh; /* dirty background threshold */ + unsigned long limit; /* hard dirty limit */ unsigned long wb_dirty; /* per-wb counterparts */ unsigned long wb_thresh; diff --git a/include/trace/events/writeback.h b/include/trace/events/writeback.h index 3046ca6b08ea..0ff388131fc9 100644 --- a/include/trace/events/writeback.h +++ b/include/trace/events/writeback.h @@ -663,9 +663,8 @@ TRACE_EVENT(balance_dirty_pages, unsigned long freerun = (dtc->thresh + dtc->bg_thresh) / 2; strscpy_pad(__entry->bdi, bdi_dev_name(wb->bdi), 32); - __entry->limit = global_wb_domain.dirty_limit; - __entry->setpoint = (global_wb_domain.dirty_limit + - freerun) / 2; + __entry->limit = dtc->limit; + __entry->setpoint = (dtc->limit + freerun) / 2; __entry->dirty = dtc->dirty; __entry->wb_setpoint = __entry->setpoint * dtc->wb_thresh / (dtc->thresh + 1); diff --git a/mm/page-writeback.c b/mm/page-writeback.c index 149f8b815904..18456ddd463b 100644 --- a/mm/page-writeback.c +++ b/mm/page-writeback.c @@ -1072,7 +1072,7 @@ static void wb_position_ratio(struct dirty_throttle_control *dtc) struct bdi_writeback *wb = dtc->wb; unsigned long write_bw = READ_ONCE(wb->avg_write_bandwidth); unsigned long freerun = dirty_freerun_ceiling(dtc->thresh, dtc->bg_thresh); - unsigned long limit = hard_dirty_limit(dtc_dom(dtc), dtc->thresh); + unsigned long limit = dtc->limit = hard_dirty_limit(dtc_dom(dtc), dtc->thresh); unsigned long wb_thresh = dtc->wb_thresh; unsigned long x_intercept; unsigned long setpoint; /* dirty pages' target balance point */ -- cgit v1.2.3 From 8c02048d1c6126527f15752a5e0849dc49cefeeb Mon Sep 17 00:00:00 2001 From: Martin Liu Date: Sat, 8 Mar 2025 03:46:00 +0000 Subject: mm/page_alloc: add trace event for per-zone watermark setup Patch series "Add tracepoints for lowmem reserves, watermarks and totalreserve_pages", v2. This patchset introduces tracepoints to track changes in the lowmem reserves, watermarks and totalreserve_pages. This helps to track the exact timing of such changes and understand their relation to reclaim activities. The tracepoints added are: mm_setup_per_zone_lowmem_reserve mm_setup_per_zone_wmarks mm_calculate_totalreserve_pagesi This patch (of 3): This commit introduces the `mm_setup_per_zone_wmarks` trace event, which provides detailed insights into the kernel's per-zone watermark configuration, offering precise timing and the ability to correlate watermark changes with specific kernel events. While `/proc/zoneinfo` provides some information about zone watermarks, this trace event offers: 1. The ability to link watermark changes to specific kernel events and logic. 2. The ability to capture rapid or short-lived changes in watermarks that may be missed by user-space polling 3. Diagnosing unexpected kswapd activity or excessive direct reclaim triggered by rapidly changing watermarks. Link: https://lkml.kernel.org/r/20250308034606.2036033-1-liumartin@google.com Link: https://lkml.kernel.org/r/20250308034606.2036033-2-liumartin@google.com Signed-off-by: Martin Liu Acked-by: David Rientjes Cc: Steven Rostedt Cc: Martin Liu Cc: "Masami Hiramatsu (Google)" Cc: Mathieu Desnoyers Signed-off-by: Andrew Morton --- include/trace/events/kmem.h | 33 +++++++++++++++++++++++++++++++++ mm/page_alloc.c | 1 + 2 files changed, 34 insertions(+) (limited to 'include/trace') diff --git a/include/trace/events/kmem.h b/include/trace/events/kmem.h index b37eb0a7060f..5fd392dae503 100644 --- a/include/trace/events/kmem.h +++ b/include/trace/events/kmem.h @@ -342,6 +342,39 @@ TRACE_EVENT(mm_alloc_contig_migrate_range_info, __entry->nr_mapped) ); +TRACE_EVENT(mm_setup_per_zone_wmarks, + + TP_PROTO(struct zone *zone), + + TP_ARGS(zone), + + TP_STRUCT__entry( + __field(int, node_id) + __string(name, zone->name) + __field(unsigned long, watermark_min) + __field(unsigned long, watermark_low) + __field(unsigned long, watermark_high) + __field(unsigned long, watermark_promo) + ), + + TP_fast_assign( + __entry->node_id = zone->zone_pgdat->node_id; + __assign_str(name); + __entry->watermark_min = zone->_watermark[WMARK_MIN]; + __entry->watermark_low = zone->_watermark[WMARK_LOW]; + __entry->watermark_high = zone->_watermark[WMARK_HIGH]; + __entry->watermark_promo = zone->_watermark[WMARK_PROMO]; + ), + + TP_printk("node_id=%d zone name=%s watermark min=%lu low=%lu high=%lu promo=%lu", + __entry->node_id, + __get_str(name), + __entry->watermark_min, + __entry->watermark_low, + __entry->watermark_high, + __entry->watermark_promo) +); + /* * Required for uniquely and securely identifying mm in rss_stat tracepoint. */ diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 2c6ae7e5aaad..b739367434ce 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -6006,6 +6006,7 @@ static void __setup_per_zone_wmarks(void) zone->_watermark[WMARK_LOW] = min_wmark_pages(zone) + tmp; zone->_watermark[WMARK_HIGH] = low_wmark_pages(zone) + tmp; zone->_watermark[WMARK_PROMO] = high_wmark_pages(zone) + tmp; + trace_mm_setup_per_zone_wmarks(zone); spin_unlock_irqrestore(&zone->lock, flags); } -- cgit v1.2.3 From a293aba4a584709889f77a0ad0c45746aecf1b9f Mon Sep 17 00:00:00 2001 From: Martin Liu Date: Sat, 8 Mar 2025 03:46:01 +0000 Subject: mm/page_alloc: add trace event for per-zone lowmem reserve setup This commit introduces the `mm_setup_per_zone_lowmem_reserve` trace event,which provides detailed insights into the kernel's per-zone lowmem reserve configuration. The trace event provides precise timestamps, allowing developers to 1. Correlate lowmem reserve changes with specific kernel events and able to diagnose unexpected kswapd or direct reclaim behavior triggered by dynamic changes in lowmem reserve. 2. Know memory allocation failures that occur due to insufficient lowmem reserve, by precisely correlating allocation attempts with reserve adjustments. Link: https://lkml.kernel.org/r/20250308034606.2036033-3-liumartin@google.com Signed-off-by: Martin Liu Acked-by: David Rientjes Cc: "Masami Hiramatsu (Google)" Cc: Mathieu Desnoyers Cc: Steven Rostedt Signed-off-by: Andrew Morton --- include/trace/events/kmem.h | 27 +++++++++++++++++++++++++++ mm/page_alloc.c | 2 ++ 2 files changed, 29 insertions(+) (limited to 'include/trace') diff --git a/include/trace/events/kmem.h b/include/trace/events/kmem.h index 5fd392dae503..9623e68d4d26 100644 --- a/include/trace/events/kmem.h +++ b/include/trace/events/kmem.h @@ -375,6 +375,33 @@ TRACE_EVENT(mm_setup_per_zone_wmarks, __entry->watermark_promo) ); +TRACE_EVENT(mm_setup_per_zone_lowmem_reserve, + + TP_PROTO(struct zone *zone, struct zone *upper_zone, long lowmem_reserve), + + TP_ARGS(zone, upper_zone, lowmem_reserve), + + TP_STRUCT__entry( + __field(int, node_id) + __string(name, zone->name) + __string(upper_name, upper_zone->name) + __field(long, lowmem_reserve) + ), + + TP_fast_assign( + __entry->node_id = zone->zone_pgdat->node_id; + __assign_str(name); + __assign_str(upper_name); + __entry->lowmem_reserve = lowmem_reserve; + ), + + TP_printk("node_id=%d zone name=%s upper_zone name=%s lowmem_reserve_pages=%ld", + __entry->node_id, + __get_str(name), + __get_str(upper_name), + __entry->lowmem_reserve) +); + /* * Required for uniquely and securely identifying mm in rss_stat tracepoint. */ diff --git a/mm/page_alloc.c b/mm/page_alloc.c index b739367434ce..a82d96cb3044 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -5943,6 +5943,8 @@ static void setup_per_zone_lowmem_reserve(void) zone->lowmem_reserve[j] = 0; else zone->lowmem_reserve[j] = managed_pages / ratio; + trace_mm_setup_per_zone_lowmem_reserve(zone, upper_zone, + zone->lowmem_reserve[j]); } } } -- cgit v1.2.3 From 15766485e4a51bec2dcce304c089a95550720033 Mon Sep 17 00:00:00 2001 From: Martin Liu Date: Sat, 8 Mar 2025 03:46:02 +0000 Subject: mm/page_alloc: add trace event for totalreserve_pages calculation This commit introduces a new trace event, `mm_calculate_totalreserve_pages`, which reports the new reserve value at the exact time when it takes effect. The `totalreserve_pages` value represents the total amount of memory reserved across all zones and nodes in the system. This reserved memory is crucial for ensuring that critical kernel operations have access to sufficient memory, even under memory pressure. By tracing the `totalreserve_pages` value, developers can gain insights that how the total reserved memory changes over time. Link: https://lkml.kernel.org/r/20250308034606.2036033-4-liumartin@google.com Signed-off-by: Martin Liu Acked-by: David Rientjes Cc: "Masami Hiramatsu (Google)" Cc: Mathieu Desnoyers Cc: Steven Rostedt Signed-off-by: Andrew Morton --- include/trace/events/kmem.h | 18 ++++++++++++++++++ mm/page_alloc.c | 1 + 2 files changed, 19 insertions(+) (limited to 'include/trace') diff --git a/include/trace/events/kmem.h b/include/trace/events/kmem.h index 9623e68d4d26..f74925a6cf69 100644 --- a/include/trace/events/kmem.h +++ b/include/trace/events/kmem.h @@ -402,6 +402,24 @@ TRACE_EVENT(mm_setup_per_zone_lowmem_reserve, __entry->lowmem_reserve) ); +TRACE_EVENT(mm_calculate_totalreserve_pages, + + TP_PROTO(unsigned long totalreserve_pages), + + TP_ARGS(totalreserve_pages), + + TP_STRUCT__entry( + __field(unsigned long, totalreserve_pages) + ), + + TP_fast_assign( + __entry->totalreserve_pages = totalreserve_pages; + ), + + TP_printk("totalreserve_pages=%lu", __entry->totalreserve_pages) +); + + /* * Required for uniquely and securely identifying mm in rss_stat tracepoint. */ diff --git a/mm/page_alloc.c b/mm/page_alloc.c index a82d96cb3044..0be1fedd1201 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -5914,6 +5914,7 @@ static void calculate_totalreserve_pages(void) } } totalreserve_pages = reserve_pages; + trace_mm_calculate_totalreserve_pages(totalreserve_pages); } /* -- cgit v1.2.3