zram: Introduce an aged idle interface
Subsystems (4)
PATCH 1/1 zram: Introduce an aged idle interface
This change introduces a new sysfs file for the zram idle interface. The current idle interface has only a single mode "all." This is severely limiting in that it requires that you have the foresight to know that you'll want to have pages idle at some point in the future and forces the user to mark everything as idle.
This new proposed file idle_aged takes a single integer argument which represents the age of the pages (in seconds since access) to mark as idle. Because it requires accessed tracking it is only enabled when CONFIG_ZRAM_MEMORY_TRACKING is enabled. There are a few
reasons why this is being proposed as a new file rather than extending the existing file. The first reason is that it wouldn't allow more code sharing as this change already refactors the existing code to do so. Secondly, having a standalone file allows a caller to quickly check if this idle_aged interface is supported. Finally, it simplifies the parsing logic because now you only need to parse an int rather than more complex logic which would need to parse things like "aged 50."
--- Documentation/admin-guide/blockdev/zram.rst | 11 ++- drivers/block/zram/zram_drv.c | 75 +++++++++++++++++---- 2 files changed, 70 insertions(+), 16 deletions(-)
Documentation/admin-guide/blockdev/zram.rst +9 -2
@@ -209,6 +209,8 @@ compact WO trigger memory compaction debug_stat RO this file is used for zram debugging purposes backing_dev RW set up backend storage for zram to write out idle WO mark allocated slot as idle +idle_aged WO mark allocated slot older than 'age' seconds + as idle (see later) ====================== ====== ===============================================
@@ -325,8 +327,13 @@ as idle:: echo all > /sys/block/zramX/idle -From now on, any pages on zram are idle pages. The idle mark -will be removed until someone requests access of the block. +Alternatively if the config option CONFIG_ZRAM_MEMORY_TRACKING is enabled +the idle_aged interface can be used to mark only pages older than 'age' +seconds as idle:: + + echo 86400 > /sys/block/zramX/idle_aged + +The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. Admin can request writeback of those idle pages at right timing via::
drivers/block/zram/zram_drv.c +61 -14
111 lines not on this page · raw message
Signature · 1 line
2.33.0.464.g1972c5931b-goog
Minchan Kim
maintainerQuoted 6 lines · Brian Geffon
This change introduces a new sysfs file for the zram idle interface. The current idle interface has only a single mode "all." This is severely limiting in that it requires that you have the foresight to know that you'll want to have pages idle at some point in the future and forces the user to mark everything as idle.
Quoted 11 lines
This new proposed file
idle_agedtakes a single integer argument which represents the age of the pages (in seconds since access) to mark as idle. Because it requires accessed tracking it is only enabled whenCONFIG_ZRAM_MEMORY_TRACKINGis enabled. There are a few
reasons why this is being proposed as a new file rather than extending the existing file. The first reason is that it wouldn't allow more code sharing as this change already refactors the existing code to do so. Secondly, having a standalone file allows a caller to quickly check if thisidle_agedinterface is supported. Finally, it simplifies the parsing logic because now you only need to parse an int rather than more complex logic which would need to parse things like "aged 50."
I am not convinced with create new sysfs node to support the aged time. The reason I used the "all" was to have the room to support the aged time as you suggested for the future. IOW, If the value is not "all" but "only numeric", we could take it as "aged" value.
And then, we can write up in the doc "The age-based idle" supported by kernel version 5.XX with CONFIG_ZRAM_MEMORY_TRACKING.
I understand the separate interface will make code simple but not sure it's worth to create another sysfs knob.
Any thoughts?
Quoted 70 lines
Signed-off-by Brian Geffon <[email protected]>--- Documentation/admin-guide/blockdev/zram.rst | 11 ++- drivers/block/zram/zram_drv.c | 75 +++++++++++++++++---- 2 files changed, 70 insertions(+), 16 deletions(-)Documentation/admin-guide/blockdev/zram.rst +9
-2@@ -209,6 +209,8 @@ compact WO trigger memory compaction debug_stat RO this file is used for zram debugging purposes backing_dev RW set up backend storage for zram to write out idle WO mark allocated slot as idle +idle_aged WO mark allocated slot older than 'age' seconds + as idle (see later) ====================== ====== ===============================================@@ -325,8 +327,13 @@ as idle:: echo all > /sys/block/zramX/idle -From now on, any pages on zram are idle pages. The idle mark -will be removed until someone requests access of the block. +Alternatively if the config option CONFIG_ZRAM_MEMORY_TRACKING is enabled +the idle_aged interface can be used to mark only pages older than 'age' +seconds as idle:: + + echo 86400 > /sys/block/zramX/idle_aged + +The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. Admin can request writeback of those idle pages at right timing via::drivers/block/zram/zram_drv.c +6
-12@@ -291,22 +291,16 @@ static ssize_t mem_used_max_store(struct device *dev,return len; } -static ssize_t idle_store (struct device *dev, -struct device_attribute *attr,const char *buf,size_t len) +/* +* Mark all pages which are older than or equal to cutoff as IDLE. +* Callers should hold the zram init lock in read mode +**/ +static void mark_idle (struct zram *zram,ktime_t cutoff) { -struct zram *zram =dev_to_zram (dev); +int is_idle =1 ;unsigned long nr_pages = zram->disksize >>PAGE_SHIFT ;int index; -if (!sysfs_streq (buf,"all" )) -return -EINVAL ; - -down_read (&zram->init_lock); -if (!init_done (zram)) { -up_read (&zram->init_lock); -return -EINVAL ; - } -for (index =0 ; index < nr_pages; index++) {/* * Do not mark ZRAM_UNDER_WB slot as ZRAM_IDLE to close race.
drivers/block/zram/zram_drv.c static ssize_t idle_store(struct device *dev, · line 313
Quoted 8 lines
drivers/block/zram/zram_drv.c +3
-2@@ -314,16 +308,63 @@ static ssize_t idle_store(struct device *dev,*/ zram_slot_lock (zram, index);if (zram_allocated (zram, index) && - !zram_test_flag (zram, index,ZRAM_UNDER_WB )) -zram_set_flag (zram, index,ZRAM_IDLE ); + !zram_test_flag (zram, index,ZRAM_UNDER_WB )) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING + is_idle = (!cutoff || cutoff >= zram->table[index].ac_time);
I am not sure it's a good idea to compare those ktime_t directly without using the ktime API(e.g., ktime_after or ktime_before). Maybe, we can get age_sec ktime_t directly as a argument and then use the ktime API with ktime_get_boottime to compare them.
Quoted 80 lines
drivers/block/zram/zram_drv.c +52
-0+#endif +if (is_idle) +zram_set_flag (zram, index,ZRAM_IDLE ); + }zram_slot_unlock (zram, index); } +} + +static ssize_t idle_store (struct device *dev, +struct device_attribute *attr,const char *buf,size_t len) +{ +struct zram *zram =dev_to_zram (dev); + +if (!sysfs_streq (buf,"all" )) +return -EINVAL ; + +down_read (&zram->init_lock); +if (!init_done (zram)) { +up_read (&zram->init_lock); +return -EINVAL ; + } +/* Mark everything as idle */ +mark_idle (zram,0 );up_read (&zram->init_lock);return len; } +#ifdef CONFIG_ZRAM_MEMORY_TRACKING +static ssize_t idle_aged_store (struct device *dev, +struct device_attribute *attr,const char *buf,size_t len) +{ +struct zram *zram =dev_to_zram (dev); +ktime_t cutoff_time; +u64 age_sec; +ssize_t rv = -EINVAL ; + +down_read (&zram->init_lock); +if (!init_done (zram)) + goto out; + +if (kstrtoull (buf,10 , &age_sec)) + goto out; + + rv = len; + cutoff_time =ktime_sub (ktime_get_boottime (),ns_to_ktime (age_sec *NSEC_PER_SEC )); +mark_idle (zram, cutoff_time); +out: +up_read (&zram->init_lock); +return rv; +} +#endif +#ifdef CONFIG_ZRAM_WRITEBACK static ssize_t writeback_limit_enable_store (struct device *dev,struct device_attribute *attr,const char *buf,size_t len)@@ -1840,6 +1881,9 @@ static DEVICE_ATTR_WO(reset);static DEVICE_ATTR_WO (mem_limit);static DEVICE_ATTR_WO (mem_used_max);static DEVICE_ATTR_WO (idle); +#ifdef CONFIG_ZRAM_MEMORY_TRACKING +static DEVICE_ATTR_WO (idle_aged); +#endif static DEVICE_ATTR_RW (max_comp_streams);static DEVICE_ATTR_RW (comp_algorithm);#ifdef CONFIG_ZRAM_WRITEBACK @@ -1857,6 +1901,9 @@ static struct attribute *zram_disk_attrs[] = { &dev_attr_mem_limit.attr, &dev_attr_mem_used_max.attr, &dev_attr_idle.attr, +#ifdef CONFIG_ZRAM_MEMORY_TRACKING + &dev_attr_idle_aged.attr, +#endif &dev_attr_max_comp_streams.attr, &dev_attr_comp_algorithm.attr, #ifdef CONFIG_ZRAM_WRITEBACKSignature · 1 line
2.33.0.464.g1972c5931b-goog
Quoted 25 lines · Minchan Kim
Quoted 6 lines · Brian Geffon
This change introduces a new sysfs file for the zram idle interface. The current idle interface has only a single mode "all." This is severely limiting in that it requires that you have the foresight to know that you'll want to have pages idle at some point in the future and forces the user to mark everything as idle.
Quoted 11 lines
This new proposed file
idle_agedtakes a single integer argument which represents the age of the pages (in seconds since access) to mark as idle. Because it requires accessed tracking it is only enabled whenCONFIG_ZRAM_MEMORY_TRACKINGis enabled. There are a few
reasons why this is being proposed as a new file rather than extending the existing file. The first reason is that it wouldn't allow more code sharing as this change already refactors the existing code to do so. Secondly, having a standalone file allows a caller to quickly check if thisidle_agedinterface is supported. Finally, it simplifies the parsing logic because now you only need to parse an int rather than more complex logic which would need to parse things like "aged 50."I am not convinced with create new sysfs node to support the aged time. The reason I used the "all" was to have the room to support the aged time as you suggested for the future. IOW, If the value is not "all" but "only numeric", we could take it as "aged" value.
Quoted 6 lines
And then, we can write up in the doc "The age-based idle" supported by kernel version 5.XX with
CONFIG_ZRAM_MEMORY_TRACKING.
I understand the separate interface will make code simple but not sure it's worth to create another sysfs knob.Any thoughts?
Sure, Sergey had similar feedback on that. I'll mail a new patch that uses the existing knob but attempts to parse the input as an integer.
Quoted 76 lines
Quoted 76 lines
Signed-off-by Brian Geffon <[email protected]>--- Documentation/admin-guide/blockdev/zram.rst | 11 ++- drivers/block/zram/zram_drv.c | 75 +++++++++++++++++---- 2 files changed, 70 insertions(+), 16 deletions(-)Documentation/admin-guide/blockdev/zram.rst +9
-2@@ -209,6 +209,8 @@ compact WO trigger memory compaction debug_stat RO this file is used for zram debugging purposes backing_dev RW set up backend storage for zram to write out idle WO mark allocated slot as idle +idle_aged WO mark allocated slot older than 'age' seconds + as idle (see later) ====================== ====== ===============================================@@ -325,8 +327,13 @@ as idle:: echo all > /sys/block/zramX/idle -From now on, any pages on zram are idle pages. The idle mark -will be removed until someone requests access of the block. +Alternatively if the config option CONFIG_ZRAM_MEMORY_TRACKING is enabled +the idle_aged interface can be used to mark only pages older than 'age' +seconds as idle:: + + echo 86400 > /sys/block/zramX/idle_aged + +The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. Admin can request writeback of those idle pages at right timing via::drivers/block/zram/zram_drv.c +6
-14@@ -291,22 +291,16 @@ static ssize_t mem_used_max_store(struct device *dev,return len; } -static ssize_t idle_store (struct device *dev, -struct device_attribute *attr,const char *buf,size_t len) +/* +* Mark all pages which are older than or equal to cutoff as IDLE. +* Callers should hold the zram init lock in read mode +**/ +static void mark_idle (struct zram *zram,ktime_t cutoff) { -struct zram *zram =dev_to_zram (dev); +int is_idle =1 ;unsigned long nr_pages = zram->disksize >>PAGE_SHIFT ;int index; -if (!sysfs_streq (buf,"all" )) -return -EINVAL ; - -down_read (&zram->init_lock); -if (!init_done (zram)) { -up_read (&zram->init_lock); -return -EINVAL ; - } -for (index =0 ; index < nr_pages; index++) {/* * Do not mark ZRAM_UNDER_WB slot as ZRAM_IDLE to close race. @@ -314,16 +308,63 @@ static ssize_t idle_store(struct device *dev,*/ zram_slot_lock (zram, index); if (zram_allocated(zram, index) && - !zram_test_flag (zram, index,ZRAM_UNDER_WB )) -zram_set_flag (zram, index,ZRAM_IDLE );
Quoted 8 lines
drivers/block/zram/zram_drv.c static ssize_t idle_store(struct device *dev, · line 313
Quoted 3 lines
drivers/block/zram/zram_drv.c +3
-0+ !zram_test_flag(zram, index, ZRAM_UNDER_WB)) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING + is_idle = (!cutoff || cutoff >= zram->table[index].ac_time);I am not sure it's a good idea to compare those
ktime_tdirectly without using the ktimeAPI(e.g., ktime_after or ktime_before). Maybe, we can getage_secktime_tdirectly as a argument and then use the ktime API withktime_get_boottimeto compare them.
I will make this change in the next version of the patch.
Quoted 73 lines
Quoted 73 lines
drivers/block/zram/zram_drv.c +50
-0+#endif +if (is_idle) +zram_set_flag (zram, index,ZRAM_IDLE ); + }zram_slot_unlock (zram, index); } +} + +static ssize_t idle_store (struct device *dev, +struct device_attribute *attr,const char *buf,size_t len) +{ +struct zram *zram =dev_to_zram (dev); + +if (!sysfs_streq (buf,"all" )) +return -EINVAL ; + +down_read (&zram->init_lock); +if (!init_done (zram)) { +up_read (&zram->init_lock); +return -EINVAL ; + } +/* Mark everything as idle */ +mark_idle (zram,0 );up_read (&zram->init_lock);return len; } +#ifdef CONFIG_ZRAM_MEMORY_TRACKING +static ssize_t idle_aged_store (struct device *dev, +struct device_attribute *attr,const char *buf,size_t len) +{ +struct zram *zram =dev_to_zram (dev); +ktime_t cutoff_time; +u64 age_sec; +ssize_t rv = -EINVAL ; + +down_read (&zram->init_lock); +if (!init_done (zram)) + goto out; + +if (kstrtoull (buf,10 , &age_sec)) + goto out; + + rv = len; + cutoff_time =ktime_sub (ktime_get_boottime (),ns_to_ktime (age_sec *NSEC_PER_SEC )); +mark_idle (zram, cutoff_time); +out: +up_read (&zram->init_lock); +return rv; +} +#endif +#ifdef CONFIG_ZRAM_WRITEBACK static ssize_t writeback_limit_enable_store (struct device *dev,struct device_attribute *attr,const char *buf,size_t len)@@ -1840,6 +1881,9 @@ static DEVICE_ATTR_WO(reset);static DEVICE_ATTR_WO (mem_limit);static DEVICE_ATTR_WO (mem_used_max);static DEVICE_ATTR_WO (idle); +#ifdef CONFIG_ZRAM_MEMORY_TRACKING +static DEVICE_ATTR_WO (idle_aged); +#endif static DEVICE_ATTR_RW (max_comp_streams);static DEVICE_ATTR_RW (comp_algorithm);#ifdef CONFIG_ZRAM_WRITEBACK @@ -1857,6 +1901,9 @@ static struct attribute *zram_disk_attrs[] = { &dev_attr_mem_limit.attr, &dev_attr_mem_used_max.attr, &dev_attr_idle.attr, +#ifdef CONFIG_ZRAM_MEMORY_TRACKING
Quoted 8 lines
Quoted 8 lines
drivers/block/zram/zram_drv.c +2
-1+ &dev_attr_idle_aged.attr, +#endif &dev_attr_max_comp_streams.attr, &dev_attr_comp_algorithm.attr, #ifdef CONFIG_ZRAM_WRITEBACK --2.33.0.464.g1972c5931b-goog
Thank you for looking,
Brian
PATCH v2 1/1 zram: Introduce an aged idle interface
In reply to Brian GeffonThis change introduces an aged idle interface to the existing idle sysfs file for zram.
When CONFIG_ZRAM_MEMORY_TRACKING is enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.
--- Documentation/admin-guide/blockdev/zram.rst | 8 +++ drivers/block/zram/zram_drv.c | 60 +++++++++++++++------ 2 files changed, 52 insertions(+), 16 deletions(-)
Documentation/admin-guide/blockdev/zram.rst +8 -0
@@ -328,6 +328,14 @@ as idle:: From now on, any pages on zram are idle pages. The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. +Additionally, when CONFIG_ZRAM_MEMORY_TRACKING is enabled pages can be +marked as idle based on how long (in seconds) it's been since they were +last accessed, in seconds:: + + echo 86400 > /sys/block/zramX/idle + +In this example all pages which haven't been accessed in more than 86400 +seconds (one day) will be marked idle. Admin can request writeback of those idle pages at right timing via::
drivers/block/zram/zram_drv.c +44 -16
80 lines not on this page · raw message
Signature · 1 line
2.33.0.464.g1972c5931b-goog
PATCH v3 1/1 zram: Introduce an aged idle interface
In reply to Brian GeffonThis change introduces an aged idle interface to the existing idle sysfs file for zram.
When CONFIG_ZRAM_MEMORY_TRACKING is enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.
v2 -> v3:
- Correct unused variable warning when CONFIG_ZRAM_MEMORY_TRACKING is not enabled.
v1 -> v2:
- Switch to using existing idle file.
- Dont compare ktime directly.
--- Documentation/admin-guide/blockdev/zram.rst | 8 +++ drivers/block/zram/zram_drv.c | 60 +++++++++++++++------ 2 files changed, 52 insertions(+), 16 deletions(-)
Documentation/admin-guide/blockdev/zram.rst +8 -0
@@ -328,6 +328,14 @@ as idle:: From now on, any pages on zram are idle pages. The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. +Additionally, when CONFIG_ZRAM_MEMORY_TRACKING is enabled pages can be +marked as idle based on how long (in seconds) it's been since they were +last accessed, in seconds:: + + echo 86400 > /sys/block/zramX/idle + +In this example all pages which haven't been accessed in more than 86400 +seconds (one day) will be marked idle. Admin can request writeback of those idle pages at right timing via::
drivers/block/zram/zram_drv.c +44 -16
80 lines not on this page · raw message
Signature · 1 line
2.33.0.464.g1972c5931b-goog
Minchan Kim
maintainerQuoted 96 lines · Brian Geffon
This change introduces an aged idle interface to the existing idle sysfs file for zram.
When
CONFIG_ZRAM_MEMORY_TRACKINGis enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.v2 -> v3:
- Correct unused variable warning whenCONFIG_ZRAM_MEMORY_TRACKINGis not enabled.
v1 -> v2:
- Switch to using existing idle file.
- Dont compare ktime directly.Signed-off-by Brian Geffon <[email protected]>--- Documentation/admin-guide/blockdev/zram.rst | 8 +++ drivers/block/zram/zram_drv.c | 60 +++++++++++++++------ 2 files changed, 52 insertions(+), 16 deletions(-)Documentation/admin-guide/blockdev/zram.rst +8
-0@@ -328,6 +328,14 @@ as idle:: From now on, any pages on zram are idle pages. The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. +Additionally, when CONFIG_ZRAM_MEMORY_TRACKING is enabled pages can be +marked as idle based on how long (in seconds) it's been since they were +last accessed, in seconds:: + + echo 86400 > /sys/block/zramX/idle + +In this example all pages which haven't been accessed in more than 86400 +seconds (one day) will be marked idle. Admin can request writeback of those idle pages at right timing via::drivers/block/zram/zram_drv.c +17
-15@@ -291,22 +291,16 @@ static ssize_t mem_used_max_store(struct device *dev,return len; } -static ssize_t idle_store (struct device *dev, -struct device_attribute *attr,const char *buf,size_t len) +/* +* Mark all pages which are older than or equal to cutoff as IDLE. +* Callers should hold the zram init lock in read mode +**/ +static void mark_idle (struct zram *zram,ktime_t cutoff) { -struct zram *zram =dev_to_zram (dev); +int is_idle =1 ;unsigned long nr_pages = zram->disksize >>PAGE_SHIFT ;int index; -if (!sysfs_streq (buf,"all" )) -return -EINVAL ; - -down_read (&zram->init_lock); -if (!init_done (zram)) { -up_read (&zram->init_lock); -return -EINVAL ; - } -for (index =0 ; index < nr_pages; index++) {/* * Do not mark ZRAM_UNDER_WB slot as ZRAM_IDLE to close race. @@ -314,14 +308,48 @@ static ssize_t idle_store(struct device *dev,*/ zram_slot_lock (zram, index);if (zram_allocated (zram, index) && - !zram_test_flag (zram, index,ZRAM_UNDER_WB )) -zram_set_flag (zram, index,ZRAM_IDLE ); + !zram_test_flag (zram, index,ZRAM_UNDER_WB )) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING + is_idle = (!cutoff ||ktime_after (cutoff, zram->table[index].ac_time)); +#endif +if (is_idle) +zram_set_flag (zram, index,ZRAM_IDLE ); + }zram_slot_unlock (zram, index); } +} -up_read (&zram->init_lock); +static ssize_t idle_store (struct device *dev, +struct device_attribute *attr,const char *buf,size_t len) +{
drivers/block/zram/zram_drv.c static ssize_t idle_store(struct device *dev, · line 333
Quoted 10 lines
drivers/block/zram/zram_drv.c +8
-1+struct zram *zram =dev_to_zram (dev); +ktime_t cutoff_time =0 ; +ssize_t rv = -EINVAL ; -return len; +if (!sysfs_streq (buf,"all" )) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING +u64 age_sec; +/* If it did not parse as 'all' try to treat it as an integer */ +if (!kstrtoull (buf,10 , &age_sec))
nit:
Do we need such limit base which work with only 10 base? Passing 0 would give more flexibility.
Otherwise, looks good to me.
Thanks, Brian.
Quoted 25 lines
drivers/block/zram/zram_drv.c +19
-0+ cutoff_time =ktime_sub (ktime_get_boottime (), +ns_to_ktime (age_sec *NSEC_PER_SEC )); + else +#endif + goto out; + } + +down_read (&zram->init_lock); +if (!init_done (zram)) + goto out_unlock; + +/* A age_sec of 0 marks everything as idle, this is the "all" behavior */ +mark_idle (zram, cutoff_time); + rv = len; + +out_unlock: +up_read (&zram->init_lock); +out: +return rv; }#ifdef CONFIG_ZRAM_WRITEBACK Signature · 1 line
2.33.0.464.g1972c5931b-goog
Hi Minchan,
Thank you for taking a look. I'm happy to make that change, but I personally cannot see why userspace would want to do something like idle pages older than "0x3C seconds" or "0o250600 seconds," it just seems like a strange way to represent seconds. What do you think?
Brian
Quoted 142 lines · Minchan Kim
Quoted 96 lines · Brian Geffon
This change introduces an aged idle interface to the existing idle sysfs file for zram.
When
CONFIG_ZRAM_MEMORY_TRACKINGis enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.v2 -> v3:
- Correct unused variable warning whenCONFIG_ZRAM_MEMORY_TRACKINGis not enabled.
v1 -> v2:
- Switch to using existing idle file.
- Dont compare ktime directly.Signed-off-by Brian Geffon <[email protected]>--- Documentation/admin-guide/blockdev/zram.rst | 8 +++ drivers/block/zram/zram_drv.c | 60 +++++++++++++++------ 2 files changed, 52 insertions(+), 16 deletions(-)Documentation/admin-guide/blockdev/zram.rst +8
-0@@ -328,6 +328,14 @@ as idle:: From now on, any pages on zram are idle pages. The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. +Additionally, when CONFIG_ZRAM_MEMORY_TRACKING is enabled pages can be +marked as idle based on how long (in seconds) it's been since they were +last accessed, in seconds:: + + echo 86400 > /sys/block/zramX/idle + +In this example all pages which haven't been accessed in more than 86400 +seconds (one day) will be marked idle. Admin can request writeback of those idle pages at right timing via::drivers/block/zram/zram_drv.c +17
-15@@ -291,22 +291,16 @@ static ssize_t mem_used_max_store(struct device *dev,return len; } -static ssize_t idle_store (struct device *dev, -struct device_attribute *attr,const char *buf,size_t len) +/* +* Mark all pages which are older than or equal to cutoff as IDLE. +* Callers should hold the zram init lock in read mode +**/ +static void mark_idle (struct zram *zram,ktime_t cutoff) { -struct zram *zram =dev_to_zram (dev); +int is_idle =1 ;unsigned long nr_pages = zram->disksize >>PAGE_SHIFT ;int index; -if (!sysfs_streq (buf,"all" )) -return -EINVAL ; - -down_read (&zram->init_lock); -if (!init_done (zram)) { -up_read (&zram->init_lock); -return -EINVAL ; - } -for (index =0 ; index < nr_pages; index++) {/* * Do not mark ZRAM_UNDER_WB slot as ZRAM_IDLE to close race. @@ -314,14 +308,48 @@ static ssize_t idle_store(struct device *dev,*/ zram_slot_lock (zram, index);if (zram_allocated (zram, index) && - !zram_test_flag (zram, index,ZRAM_UNDER_WB )) -zram_set_flag (zram, index,ZRAM_IDLE ); + !zram_test_flag (zram, index,ZRAM_UNDER_WB )) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING + is_idle = (!cutoff ||ktime_after (cutoff, zram->table[index].ac_time)); +#endif +if (is_idle) +zram_set_flag (zram, index,ZRAM_IDLE ); + }zram_slot_unlock (zram, index); } +} -up_read (&zram->init_lock); +static ssize_t idle_store (struct device *dev, +struct device_attribute *attr,const char *buf,size_t len) +{drivers/block/zram/zram_drv.c static ssize_t idle_store(struct device *dev, · line 333
Quoted 10 lines
drivers/block/zram/zram_drv.c +8
-1+struct zram *zram =dev_to_zram (dev); +ktime_t cutoff_time =0 ; +ssize_t rv = -EINVAL ; -return len; +if (!sysfs_streq (buf,"all" )) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING +u64 age_sec; +/* If it did not parse as 'all' try to treat it as an integer */ +if (!kstrtoull (buf,10 , &age_sec))nit:
Do we need such limit base which work with only 10 base? Passing 0 would give more flexibility.Otherwise, looks good to me.
Thanks, Brian.
Quoted 25 lines
drivers/block/zram/zram_drv.c +19
-1+ cutoff_time =ktime_sub (ktime_get_boottime (), +ns_to_ktime (age_sec *NSEC_PER_SEC )); + else +#endif + goto out; + } + +down_read (&zram->init_lock); +if (!init_done (zram)) + goto out_unlock; + +/* A age_sec of 0 marks everything as idle, this is the "all" behavior */ +mark_idle (zram, cutoff_time); + rv = len; + +out_unlock: +up_read (&zram->init_lock); +out: +return rv; }#ifdef CONFIG_ZRAM_WRITEBACK --2.33.0.464.g1972c5931b-goog
Minchan Kim
maintainerHey Brian,
Quoted 5 lines · Brian Geffon
Hi Minchan,
Thank you for taking a look. I'm happy to make that change, but I personally cannot see why userspace would want to do something like idle pages older than "0x3Cseconds" or "0o250600 seconds," it just seems like a strange way to represent seconds. What do you think?
Kernel communty loves inline reply instead of top posting. ;-)
I am not strong opinion about mutiple base support. The question just started from "what's the benefit with only 10-base support?" if we can support multiple bases with almost zero maintainace overhead.
Quoted 146 lines
Brian
Quoted 142 lines · Minchan Kim
Quoted 96 lines · Brian Geffon
This change introduces an aged idle interface to the existing idle sysfs file for zram.
When
CONFIG_ZRAM_MEMORY_TRACKINGis enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.v2 -> v3:
- Correct unused variable warning whenCONFIG_ZRAM_MEMORY_TRACKINGis not enabled.
v1 -> v2:
- Switch to using existing idle file.
- Dont compare ktime directly.Signed-off-by Brian Geffon <[email protected]>--- Documentation/admin-guide/blockdev/zram.rst | 8 +++ drivers/block/zram/zram_drv.c | 60 +++++++++++++++------ 2 files changed, 52 insertions(+), 16 deletions(-)Documentation/admin-guide/blockdev/zram.rst +8
-0@@ -328,6 +328,14 @@ as idle:: From now on, any pages on zram are idle pages. The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. +Additionally, when CONFIG_ZRAM_MEMORY_TRACKING is enabled pages can be +marked as idle based on how long (in seconds) it's been since they were +last accessed, in seconds:: + + echo 86400 > /sys/block/zramX/idle + +In this example all pages which haven't been accessed in more than 86400 +seconds (one day) will be marked idle. Admin can request writeback of those idle pages at right timing via::drivers/block/zram/zram_drv.c +17
-15@@ -291,22 +291,16 @@ static ssize_t mem_used_max_store(struct device *dev,return len; } -static ssize_t idle_store (struct device *dev, -struct device_attribute *attr,const char *buf,size_t len) +/* +* Mark all pages which are older than or equal to cutoff as IDLE. +* Callers should hold the zram init lock in read mode +**/ +static void mark_idle (struct zram *zram,ktime_t cutoff) { -struct zram *zram =dev_to_zram (dev); +int is_idle =1 ;unsigned long nr_pages = zram->disksize >>PAGE_SHIFT ;int index; -if (!sysfs_streq (buf,"all" )) -return -EINVAL ; - -down_read (&zram->init_lock); -if (!init_done (zram)) { -up_read (&zram->init_lock); -return -EINVAL ; - } -for (index =0 ; index < nr_pages; index++) {/* * Do not mark ZRAM_UNDER_WB slot as ZRAM_IDLE to close race. @@ -314,14 +308,48 @@ static ssize_t idle_store(struct device *dev,*/ zram_slot_lock (zram, index);if (zram_allocated (zram, index) && - !zram_test_flag (zram, index,ZRAM_UNDER_WB )) -zram_set_flag (zram, index,ZRAM_IDLE ); + !zram_test_flag (zram, index,ZRAM_UNDER_WB )) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING + is_idle = (!cutoff ||ktime_after (cutoff, zram->table[index].ac_time)); +#endif +if (is_idle) +zram_set_flag (zram, index,ZRAM_IDLE ); + }zram_slot_unlock (zram, index); } +} -up_read (&zram->init_lock); +static ssize_t idle_store (struct device *dev, +struct device_attribute *attr,const char *buf,size_t len) +{drivers/block/zram/zram_drv.c static ssize_t idle_store(struct device *dev, · line 333
Quoted 10 lines
drivers/block/zram/zram_drv.c +8
-1+struct zram *zram =dev_to_zram (dev); +ktime_t cutoff_time =0 ; +ssize_t rv = -EINVAL ; -return len; +if (!sysfs_streq (buf,"all" )) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING +u64 age_sec; +/* If it did not parse as 'all' try to treat it as an integer */ +if (!kstrtoull (buf,10 , &age_sec))nit:
Do we need such limit base which work with only 10 base? Passing 0 would give more flexibility.Otherwise, looks good to me.
Thanks, Brian.
Quoted 25 lines
drivers/block/zram/zram_drv.c +19
-1+ cutoff_time =ktime_sub (ktime_get_boottime (), +ns_to_ktime (age_sec *NSEC_PER_SEC )); + else +#endif + goto out; + } + +down_read (&zram->init_lock); +if (!init_done (zram)) + goto out_unlock; + +/* A age_sec of 0 marks everything as idle, this is the "all" behavior */ +mark_idle (zram, cutoff_time); + rv = len; + +out_unlock: +up_read (&zram->init_lock); +out: +return rv; }#ifdef CONFIG_ZRAM_WRITEBACK --2.33.0.464.g1972c5931b-goog
PATCH v4 1/1 zram: Introduce an aged idle interface
In reply to Brian GeffonThis change introduces an aged idle interface to the existing idle sysfs file for zram.
When CONFIG_ZRAM_MEMORY_TRACKING is enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.
v3 -> v4:
- Remove base10 restriction.
v2 -> v3:
- Correct unused variable warning when CONFIG_ZRAM_MEMORY_TRACKING is not enabled.
v1 -> v2:
- Switch to using existing idle file.
- Dont compare ktime directly.
--- Documentation/admin-guide/blockdev/zram.rst | 8 +++ drivers/block/zram/zram_drv.c | 60 +++++++++++++++------ 2 files changed, 52 insertions(+), 16 deletions(-)
Documentation/admin-guide/blockdev/zram.rst +8 -0
@@ -328,6 +328,14 @@ as idle:: From now on, any pages on zram are idle pages. The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. +Additionally, when CONFIG_ZRAM_MEMORY_TRACKING is enabled pages can be +marked as idle based on how long (in seconds) it's been since they were +last accessed:: + + echo 86400 > /sys/block/zramX/idle + +In this example all pages which haven't been accessed in more than 86400 +seconds (one day) will be marked idle. Admin can request writeback of those idle pages at right timing via::
drivers/block/zram/zram_drv.c +44 -16
80 lines not on this page · raw message
Signature · 1 line
2.33.0.464.g1972c5931b-goog
Quoted 13 lines · Brian Geffon
This change introduces an aged idle interface to the existing idle sysfs file for zram.
When
CONFIG_ZRAM_MEMORY_TRACKINGis enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.v3 -> v4:
- Remove base10 restriction.
Quoted 8 lines
v2 -> v3:
- Correct unused variable warning whenCONFIG_ZRAM_MEMORY_TRACKINGis not enabled.
v1 -> v2:
- Switch to using existing idle file.
- Dont compare ktime directly.Signed-off-by Brian Geffon <[email protected]>
Quoted 11 lines · On Thu, 23 Sep 2021 06:01:15 -0700 Brian Geffon <[email protected]> wrote:
This change introduces an aged idle interface to the existing idle sysfs file for zram.
When
CONFIG_ZRAM_MEMORY_TRACKINGis enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.
drivers/block/zram/zram_drv.c static ssize_t mem_used_max_store(struct device *dev, · line 297
Quoted 9 lines
drivers/block/zram/zram_drv.c +4
-2@@ -291,22 +291,16 @@ static ssize_t mem_used_max_store(struct device *dev,return len; } -static ssize_t idle_store (struct device *dev, -struct device_attribute *attr,const char *buf,size_t len) +/* +* Mark all pages which are older than or equal to cutoff as IDLE. +* Callers should hold the zram init lock in read mode +**/
A simple "*/" is conventional.
Quoted 46 lines
drivers/block/zram/zram_drv.c +17
-14+static void mark_idle (struct zram *zram,ktime_t cutoff) { -struct zram *zram =dev_to_zram (dev); +int is_idle =1 ;unsigned long nr_pages = zram->disksize >>PAGE_SHIFT ;int index; -if (!sysfs_streq (buf,"all" )) -return -EINVAL ; - -down_read (&zram->init_lock); -if (!init_done (zram)) { -up_read (&zram->init_lock); -return -EINVAL ; - } -for (index =0 ; index < nr_pages; index++) {/* * Do not mark ZRAM_UNDER_WB slot as ZRAM_IDLE to close race. @@ -314,14 +308,48 @@ static ssize_t idle_store(struct device *dev,*/ zram_slot_lock (zram, index);if (zram_allocated (zram, index) && - !zram_test_flag (zram, index,ZRAM_UNDER_WB )) -zram_set_flag (zram, index,ZRAM_IDLE ); + !zram_test_flag (zram, index,ZRAM_UNDER_WB )) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING + is_idle = (!cutoff ||ktime_after (cutoff, zram->table[index].ac_time)); +#endif +if (is_idle) +zram_set_flag (zram, index,ZRAM_IDLE ); + }zram_slot_unlock (zram, index); } +} -up_read (&zram->init_lock); +static ssize_t idle_store (struct device *dev, +struct device_attribute *attr,const char *buf,size_t len) +{ +struct zram *zram =dev_to_zram (dev); +ktime_t cutoff_time =0 ; +ssize_t rv = -EINVAL ; -return len; +if (!sysfs_streq (buf,"all" )) {
drivers/block/zram/zram_drv.c static ssize_t idle_store(struct device *dev, · line 339
Quoted 10 lines
drivers/block/zram/zram_drv.c +10
-0+#ifdef CONFIG_ZRAM_MEMORY_TRACKING +u64 age_sec; +/* If it did not parse as 'all' try to treat it as an integer */ +if (!kstrtoull (buf,0 , &age_sec)) + cutoff_time =ktime_sub (ktime_get_boottime (), +ns_to_ktime (age_sec *NSEC_PER_SEC )); + else +#endif + goto out; + }
The ifdef tricks are pretty ugly. Can things be improved with IS_ENABLED()?
PATCH v5 1/1 zram: Introduce an aged idle interface
In reply to Brian GeffonThis change introduces an aged idle interface to the existing idle sysfs file for zram.
When CONFIG_ZRAM_MEMORY_TRACKING is enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.
v4 -> v5:
- Andrew's suggestions to use IS_ENABLED and cleanup comment.
v3 -> v4:
- Remove base10 restriction.
v2 -> v3:
- Correct unused variable warning when CONFIG_ZRAM_MEMORY_TRACKING is not enabled.
v1 -> v2:
- Switch to using existing idle file.
- Dont compare ktime directly.
--- Documentation/admin-guide/blockdev/zram.rst | 8 +++ drivers/block/zram/zram_drv.c | 61 +++++++++++++++------ 2 files changed, 53 insertions(+), 16 deletions(-)
Documentation/admin-guide/blockdev/zram.rst +8 -0
@@ -327,6 +327,14 @@ as idle:: From now on, any pages on zram are idle pages. The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. +Additionally, when CONFIG_ZRAM_MEMORY_TRACKING is enabled pages can be +marked as idle based on how long (in seconds) it's been since they were +last accessed:: + + echo 86400 > /sys/block/zramX/idle + +In this example all pages which haven't been accessed in more than 86400 +seconds (one day) will be marked idle. Admin can request writeback of those idle pages at right timing via::
drivers/block/zram/zram_drv.c +45 -16
81 lines not on this page · raw message
Signature · 1 line
2.33.0.685.g46640cef36-goog
Quoted 10 lines · On Fri, 24 Sep 2021 09:11:28 -0700 Brian Geffon <[email protected]> wrote:
This change introduces an aged idle interface to the existing idle sysfs file for zram.
When
CONFIG_ZRAM_MEMORY_TRACKINGis enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.
Quoted 3 lines
v4 -> v5:
- Andrew's suggestions to useIS_ENABLEDand cleanup comment.
Also this?
drivers/block/zram/zram_drv.c renamed from drivers/block/zram/zram_drv.c~zram-introduce-an-aged-idle-interface-v5-fix +1 -2
@@ -309,9 +309,8 @@ static void mark_idle(struct zram *zram,zram_slot_lock (zram, index);if (zram_allocated (zram, index) && !zram_test_flag (zram, index,ZRAM_UNDER_WB )) { -#ifdef CONFIG_ZRAM_MEMORY_TRACKING +if (IS_ENABLED (CONFIG_ZRAM_MEMORY_TRACKING )) is_idle = (!cutoff ||ktime_after (cutoff, zram->table[index].ac_time)); -#endif if (is_idle)zram_set_flag (zram, index,ZRAM_IDLE ); }
_
Quoted 15 lines
Also this?
drivers/block/zram/zram_drv.c renamed from drivers/block/zram/zram_drv.c~zram-introduce-an-aged-idle-interface-v5-fix +1
-2@@ -309,9 +309,8 @@ static void mark_idle(struct zram *zram,zram_slot_lock (zram, index);if (zram_allocated (zram, index) && !zram_test_flag (zram, index,ZRAM_UNDER_WB )) { -#ifdef CONFIG_ZRAM_MEMORY_TRACKING +if (IS_ENABLED (CONFIG_ZRAM_MEMORY_TRACKING )) is_idle = (!cutoff ||ktime_after (cutoff, zram->table[index].ac_time)); -#endif if (is_idle)zram_set_flag (zram, index,ZRAM_IDLE ); }
Quoted 2 lines
_
Hi Andrew,
As written that patch won't compile when CONFIG_ZRAM_MEMORY_TRACKING=n, my guess is that the compiler pass that removes the dead branch only happens after it attempts to compile the branch itself. So it appears that even though IS_ENABLED(..) always evaluates to 0, the compile will fail because table[index].ac_time does not exist. You should get an error like this:
drivers/block/zram/zram_drv.c:314:57: error: no member named 'ac_time' in 'struct zram_table_entry'
(!cutoff
|| ktime_after(cutoff, zram->table[index].ac_time)))
Brian
Sergey Senozhatsky
maintainerOn (21/09/24 09:11), Brian Geffon wrote:
[..]
Some silly nits:
Quoted 20 lines
drivers/block/zram/zram_drv.c +2
-10+static void mark_idle (struct zram *zram,ktime_t cutoff) { -struct zram *zram =dev_to_zram (dev); +int is_idle =1 ;unsigned long nr_pages = zram->disksize >>PAGE_SHIFT ;int index; -if (!sysfs_streq (buf,"all" )) -return -EINVAL ; - -down_read (&zram->init_lock); -if (!init_done (zram)) { -up_read (&zram->init_lock); -return -EINVAL ; - } -for (index =0 ; index < nr_pages; index++) {/* * Do not mark ZRAM_UNDER_WB slot as ZRAM_IDLE to close race.
drivers/block/zram/zram_drv.c static ssize_t idle_store(struct device *dev, · line 313
Quoted 8 lines
drivers/block/zram/zram_drv.c +3
-2@@ -314,14 +308,49 @@ static ssize_t idle_store(struct device *dev,*/ zram_slot_lock (zram, index);if (zram_allocated (zram, index) && - !zram_test_flag (zram, index,ZRAM_UNDER_WB )) -zram_set_flag (zram, index,ZRAM_IDLE ); + !zram_test_flag (zram, index,ZRAM_UNDER_WB )) { +#ifdef CONFIG_ZRAM_MEMORY_TRACKING + is_idle = (!cutoff ||ktime_after (cutoff, zram->table[index].ac_time));
checkpatch "WARNING: suspect code indent for conditional statements (16, 32)"
Looks like is_idle is at one extra indent level.
Quoted 13 lines
drivers/block/zram/zram_drv.c +9
-1+#endif +if (is_idle) +zram_set_flag (zram, index,ZRAM_IDLE ); + }zram_slot_unlock (zram, index); } +} -up_read (&zram->init_lock); +static ssize_t idle_store (struct device *dev, +struct device_attribute *attr,const char *buf,size_t len) +{ +struct zram *zram =dev_to_zram (dev);
drivers/block/zram/zram_drv.c static ssize_t idle_store(struct device *dev, · line 334
Quoted 10 lines
drivers/block/zram/zram_drv.c +8
-1+ktime_t cutoff_time =0 ; +ssize_t rv = -EINVAL ; -return len; +if (!sysfs_streq (buf,"all" )) { +/* +* If it did not parse as 'all' try to treat it as an integer when +* we have memory tracking enabled. +*/ +u64 age_sec;
checkpatch "WARNING: Missing a blank line after declarations"
drivers/block/zram/zram_drv.c static ssize_t idle_store(struct device *dev, · line 346
Quoted 12 lines
drivers/block/zram/zram_drv.c +12
-0+if (IS_ENABLED (CONFIG_ZRAM_MEMORY_TRACKING ) && !kstrtoull (buf,0 , &age_sec)) + cutoff_time =ktime_sub (ktime_get_boottime (), +ns_to_ktime (age_sec *NSEC_PER_SEC )); +else + goto out; + } + +down_read (&zram->init_lock); +if (!init_done (zram)) + goto out_unlock; + +/* A age_sec of 0 marks everything as idle, this is the "all" behavior */
s/age_sec/cutoff_time/
Quoted 2 lines
drivers/block/zram/zram_drv.c +2
-0+mark_idle (zram, cutoff_time); + rv = len;
PATCH v6 1/1 zram: Introduce an aged idle interface
In reply to Brian GeffonThis change introduces an aged idle interface to the existing idle sysfs file for zram.
When CONFIG_ZRAM_MEMORY_TRACKING is enabled the idle file now also accepts an integer argument. This integer is the age (in seconds) of pages to mark as idle. The idle file still supports 'all' as it always has. This new approach allows for much more control over which pages get marked as idle.
v5 -> v6:
- Sergey's cleanup suggestions.
v4 -> v5:
- Andrew's suggestions to use IS_ENABLED and cleanup comment.
v3 -> v4:
- Remove base10 restriction.
v2 -> v3:
- Correct unused variable warning when CONFIG_ZRAM_MEMORY_TRACKING is not enabled.
v1 -> v2:
- Switch to using existing idle file.
- Dont compare ktime directly.
--- Documentation/admin-guide/blockdev/zram.rst | 8 +++ drivers/block/zram/zram_drv.c | 62 +++++++++++++++------ 2 files changed, 54 insertions(+), 16 deletions(-)
Documentation/admin-guide/blockdev/zram.rst +8 -0
@@ -328,6 +328,14 @@ as idle:: From now on, any pages on zram are idle pages. The idle mark will be removed until someone requests access of the block. IOW, unless there is access request, those pages are still idle pages. +Additionally, when CONFIG_ZRAM_MEMORY_TRACKING is enabled pages can be +marked as idle based on how long (in seconds) it's been since they were +last accessed:: + + echo 86400 > /sys/block/zramX/idle + +In this example all pages which haven't been accessed in more than 86400 +seconds (one day) will be marked idle. Admin can request writeback of those idle pages at right timing via::
drivers/block/zram/zram_drv.c +46 -16
82 lines not on this page · raw message
Signature · 1 line
2.33.0.685.g46640cef36-goog