On 10/29/2018 03:25 PM, Davidlohr Bueso wrote: > A recent report from a large database vendor which I shall not name > shows concerns about poor performance when consuming /proc/stat info. > Particularly kstat_irq() pops up in the profiles and most time is > being spent there. The overall system is under a lot of irqs and > almost 1k cores, thus this comes to little surprise. > > Granted that procfs in general is not known for its performance, > nor designed for it, for that matter. Some users, however may be able > to overcome this performance limitation, some not. Therefore it isn't > bad having a kernel option for users that don't want any hard irq info > -- and care enough about this. > > This patch introduces a new /proc/stat2 file that is identical to the > regular 'stat' except that it zeroes all hard irq statistics. The new > file is a drop in replacement to stat for users that need performance. > > The stat file is not touched, of course -- this was also previously > suggested by Waiman: > https://lore.kernel.org/lkml/1524166562-5644-1-git-send-email-longman@xxxxxxxxxx/ > > Signed-off-by: Davidlohr Bueso <dbueso@xxxxxxx> I am wondering if /proc/stat_noirqs will be a more descriptive name of the intent of this new procfs file or we should just go with the more generic stat2 name. Cheers, Longman > --- > Documentation/filesystems/proc.txt | 12 +++++++--- > fs/proc/stat.c | 45 ++++++++++++++++++++++++++++++++------ > 2 files changed, 47 insertions(+), 10 deletions(-) > > diff --git a/Documentation/filesystems/proc.txt b/Documentation/filesystems/proc.txt > index 12a5e6e693b6..563b01decb1e 100644 > --- a/Documentation/filesystems/proc.txt > +++ b/Documentation/filesystems/proc.txt > @@ -27,7 +27,7 @@ Table of Contents > 1.5 SCSI info > 1.6 Parallel port info in /proc/parport > 1.7 TTY info in /proc/tty > - 1.8 Miscellaneous kernel statistics in /proc/stat > + 1.8 Miscellaneous kernel statistics in /proc/stat and /proc/stat2 > 1.9 Ext4 file system parameters > > 2 Modifying System Parameters > @@ -140,6 +140,7 @@ Table 1-1: Process specific entries in /proc > mem Memory held by this process > root Link to the root directory of this process > stat Process status > + stat2 Process status without irq information > statm Process memory status information > status Process status in human readable form > wchan Present with CONFIG_KALLSYMS=y: it shows the kernel function > @@ -1301,8 +1302,8 @@ To see which tty's are currently in use, you can simply look into the file > unknown /dev/tty 4 1-63 console > > > -1.8 Miscellaneous kernel statistics in /proc/stat > -------------------------------------------------- > +1.8 Miscellaneous kernel statistics in /proc/stat and /proc/stat2 > +----------------------------------------------------------------- > > Various pieces of information about kernel activity are available in the > /proc/stat file. All of the numbers reported in this file are aggregates > @@ -1371,6 +1372,11 @@ of the possible system softirqs. The first column is the total of all > softirqs serviced; each subsequent column is the total for that particular > softirq. > > +The stat2 file acts as a performance alternative to /proc/stat for workloads > +and systems that care and are under heavy irq load. In order to to be completely > +compatible, /proc/stat and /proc/stat2 are identical with the exception that the > +later will show 0 for any (hard)irq-related fields. This refers particularly > +to the "intr" line and 'irq' column for that aggregate in the cpu line. > > 1.9 Ext4 file system parameters > ------------------------------- > diff --git a/fs/proc/stat.c b/fs/proc/stat.c > index 535eda7857cf..349040270003 100644 > --- a/fs/proc/stat.c > +++ b/fs/proc/stat.c > @@ -79,7 +79,7 @@ static u64 get_iowait_time(int cpu) > > #endif > > -static int show_stat(struct seq_file *p, void *v) > +static int __show_stat(struct seq_file *p, void *v, bool irq_stats) > { > int i, j; > u64 user, nice, system, idle, iowait, irq, softirq, steal; > @@ -100,13 +100,17 @@ static int show_stat(struct seq_file *p, void *v) > system += kcpustat_cpu(i).cpustat[CPUTIME_SYSTEM]; > idle += get_idle_time(i); > iowait += get_iowait_time(i); > - irq += kcpustat_cpu(i).cpustat[CPUTIME_IRQ]; > softirq += kcpustat_cpu(i).cpustat[CPUTIME_SOFTIRQ]; > steal += kcpustat_cpu(i).cpustat[CPUTIME_STEAL]; > guest += kcpustat_cpu(i).cpustat[CPUTIME_GUEST]; > guest_nice += kcpustat_cpu(i).cpustat[CPUTIME_GUEST_NICE]; > - sum += kstat_cpu_irqs_sum(i); > - sum += arch_irq_stat_cpu(i); > + > + if (irq_stats) { > + irq += kcpustat_cpu(i).cpustat[CPUTIME_IRQ]; > + > + sum += kstat_cpu_irqs_sum(i); > + sum += arch_irq_stat_cpu(i); > + } > > for (j = 0; j < NR_SOFTIRQS; j++) { > unsigned int softirq_stat = kstat_softirqs_cpu(j, i); > @@ -115,7 +119,9 @@ static int show_stat(struct seq_file *p, void *v) > sum_softirq += softirq_stat; > } > } > - sum += arch_irq_stat(); > + > + if (irq_stats) > + sum += arch_irq_stat(); > > seq_put_decimal_ull(p, "cpu ", nsec_to_clock_t(user)); > seq_put_decimal_ull(p, " ", nsec_to_clock_t(nice)); > @@ -136,7 +142,8 @@ static int show_stat(struct seq_file *p, void *v) > system = kcpustat_cpu(i).cpustat[CPUTIME_SYSTEM]; > idle = get_idle_time(i); > iowait = get_iowait_time(i); > - irq = kcpustat_cpu(i).cpustat[CPUTIME_IRQ]; > + if (irq_stats) > + irq = kcpustat_cpu(i).cpustat[CPUTIME_IRQ]; > softirq = kcpustat_cpu(i).cpustat[CPUTIME_SOFTIRQ]; > steal = kcpustat_cpu(i).cpustat[CPUTIME_STEAL]; > guest = kcpustat_cpu(i).cpustat[CPUTIME_GUEST]; > @@ -158,7 +165,7 @@ static int show_stat(struct seq_file *p, void *v) > > /* sum again ? it could be updated? */ > for_each_irq_nr(j) > - seq_put_decimal_ull(p, " ", kstat_irqs_usr(j)); > + seq_put_decimal_ull(p, " ", irq_stats ? kstat_irqs_usr(j) : 0); > > seq_printf(p, > "\nctxt %llu\n" > @@ -181,6 +188,16 @@ static int show_stat(struct seq_file *p, void *v) > return 0; > } > > +static int show_stat(struct seq_file *p, void *v) > +{ > + return __show_stat(p, v, true); > +} > + > +static int show_stat2(struct seq_file *p, void *v) > +{ > + return __show_stat(p, v, false); > +} > + > static int stat_open(struct inode *inode, struct file *file) > { > unsigned int size = 1024 + 128 * num_online_cpus(); > @@ -190,6 +207,12 @@ static int stat_open(struct inode *inode, struct file *file) > return single_open_size(file, show_stat, NULL, size); > } > > +static int stat2_open(struct inode *inode, struct file *file) > +{ > + unsigned int size = 1024 + 128 * num_online_cpus(); > + return single_open_size(file, show_stat2, NULL, size); > +} > + > static const struct file_operations proc_stat_operations = { > .open = stat_open, > .read = seq_read, > @@ -197,9 +220,17 @@ static const struct file_operations proc_stat_operations = { > .release = single_release, > }; > > +static const struct file_operations proc_stat2_operations = { > + .open = stat2_open, > + .read = seq_read, > + .llseek = seq_lseek, > + .release = single_release, > +}; > + > static int __init proc_stat_init(void) > { > proc_create("stat", 0, NULL, &proc_stat_operations); > + proc_create("stat2", 0, NULL, &proc_stat2_operations); > return 0; > } > fs_initcall(proc_stat_init);
