關鍵詞：sar、sadc、ksar、/proc/stat、/proc/cpuinfo、/proc/meminfo、/proc/diskstats。
在之前有簡單介紹過sar/ksar，最近在使用中感覺需要再深入了解一下。
ksar/sar從內核采集數據，並輸出可讀性數據。分析相關源碼，有助於知道數據來龍去脈。--------------------------------------1. sar源碼概覽
ksar顯示sar/sadc獲取的數據，並圖形化顯示。數據從內核節點，到sadc/sar轉換，再到ksar顯示。----------------------------2. ksar處理流程
對照ksar每張圖表，然后sar/sadc對應的采集轉換，再到內核每個數據項含義解析。--------------------------------------------3. ksar解讀
期望是從ksar上的圖表能對應到內核的代碼，明白這些圖表數據根源。
1. sar源碼概覽
sar作為sysstat一部分，相關的工具還包括sadc、sa1、sa2。
sa1負責收集並存儲每天系統動態信息到一個二進制的文件中，sa1是sadc所涉及的程序前端程序。通常由計划任務工具cron來調用。打開sa1文件不難看出就是調用sadc今次那個采集數據。
sa2就是調用sar命令，將當日二進制日志文件數據存儲到文本文件中。
sadc是系統動態數據收集工具，收集的數據被寫入一個二進制文件中，它是sar工具后端。
sar負責解析sadc保存的數據，並顯示出來。
當使用sar進行數據統計的時候，通過pstree `pidof sar`，可以看出sar調用了sadc。
sar───sadc
1.1 sadc信息采樣
sadc入口在sadc.c中，主要是解析參數、啟動一個interval alarm、rw_sa_stat_loop()讀取數據。
通過alarm觸發SIGALRM實現周期性讀取，SIGINT停止讀取。
int main(int argc, char **argv) { int opt = 0; char ofile[MAX_FILE_LEN], sa_dir[MAX_FILE_LEN]; int stdfd = 0, ofd = -1; int restart_mark; long count = 0; /* Get HZ */ get_HZ(); /* Compute page shift in kB */ get_kb_shift(); ofile[0] = sa_dir[0] = comment[0] = '\0'; ...-----------------------------------------------------------解析參數並進行配置。 /* Set a handler for SIGALRM */ memset(&alrm_act, 0, sizeof(alrm_act)); alrm_act.sa_handler = alarm_handler; sigaction(SIGALRM, &alrm_act, NULL); alarm(interval);------------------------------------------啟動一個alarm，超時未interval，並通過在alarm_handler()中重新起一個alarm實現周期性處理。 /* Main loop */ rw_sa_stat_loop(count, stdfd, ofd, ofile, sa_dir); #ifdef HAVE_SENSORS /* Cleanup sensors */ sensors_cleanup(); #endif /* HAVE_SENSORS */ /* Free structures */ sa_sys_free(); return 0; }
rw_sa_stat_loop()是整個sadc的核心循環，這里從個sysfs讀取信息，並提取關鍵信息，然后保存。
void rw_sa_stat_loop(long count, int stdfd, int ofd, char ofile[], char sa_dir[]) { int do_sa_rotat = 0; unsigned int save_flags; char new_ofile[MAX_FILE_LEN] = ""; struct tm rectime = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, NULL}; /* Set a handler for SIGINT */ memset(&int_act, 0, sizeof(int_act)); int_act.sa_handler = int_handler; sigaction(SIGINT, &int_act, NULL); /* Main loop */ do { reset_stats(); ... /* Read then write stats */ read_stats();--------------------------------------------------------遍歷act[]中所有的struct activity，進行采樣。 if (stdfd >= 0) { save_flags = flags; flags &= ~S_F_LOCK_FILE; write_stats(stdfd);----------------------------------------------通過標准輸出文件打印信息。 flags = save_flags; } /* If the record type was R_LAST_STATS, tag it R_STATS before writing it */ record_hdr.record_type = R_STATS; if (ofile[0]) { write_stats(ofd);------------------------------------------------將結果寫到指定文件中。 } ... fflush(stdout); if (count > 0) { count--; } if (count) { /* Wait for a signal (probably SIGALRM or SIGINT) */ pause();----------------------------------------------------------此處和alarm()配合達到周期性采樣的效果。 } if (sigint_caught)----------------------------------------------------如果收到SIGINT信號，提前終止采樣。 /* SIGINT caught: Stop now */ break; ... } while (count);------------------------------------------------------------達到總采樣數，同樣停止采樣。 /* Close file descriptors if they have actually been used */ CLOSE(stdfd); CLOSE(ofd); }
read_stats()是核心采集數據函數，核心數據結構式act[]。
void read_stats(void) { int i; __nr_t cpu_nr = act[get_activity_position(act, A_CPU, EXIT_IF_NOT_FOUND)]->nr; record_hdr.uptime0 = 0; if (cpu_nr > 2) { read_uptime(&(record_hdr.uptime0)); } for (i = 0; i < NR_ACT; i++) { if (IS_COLLECTED(act[i]->options)) {---------------------------------------遍歷所有act[]，如果act[]中對應的options包含AO_COLLECTED則進行f_read()。 /* Read statistics for current activity */ (*act[i]->f_read)(act[i]); } } if (cpu_nr == 1) { record_hdr.uptime0 = record_hdr.uptime; } }
act[]存放了所有統計事件的struct activity。
struct activity *act[NR_ACT] = { &cpu_act, &pcsw_act, &irq_act, &swap_act, &paging_act, &io_act, &memory_act, &huge_act, &ktables_act, &queue_act, &serial_act, &disk_act, /* <network> */ &net_dev_act, &net_edev_act, &net_nfs_act, &net_nfsd_act, &net_sock_act, &net_ip_act, &net_eip_act, &net_icmp_act, &net_eicmp_act, &net_tcp_act, &net_etcp_act, &net_udp_act, &net_sock6_act, &net_ip6_act, &net_eip6_act, &net_icmp6_act, &net_eicmp6_act, &net_udp6_act, &fchost_act, &softnet_act, /* AO_CLOSE_MARKUP */ /* </network> */ /* <power-management> */ &pwr_cpufreq_act, &pwr_fan_act, &pwr_temp_act, &pwr_in_act, &pwr_wghfreq_act, &pwr_usb_act, /* AO_CLOSE_MARKUP */ /* </power-management> */ &filesystem_act };
1.2 sar顯示統計信息
read_sadc_stat_bunch()讀取統計信息，write_stats()調用每個struct activity的f_print()函數，write_stats_avg()調用每個struct activity的f_print_avg()函數。
f_print()和f_print_avg()或從文件中解析字符串，或啟動sadc采樣，然后再解析。
2 ksar處理流程
2.1 ksar介紹
ksar資源：ksar安裝文件。
ksar中看到的圖標是結果，這些數據是通過sadc采集，sar解析出來的。
sadc是通過讀取sysfs/procfs節點來獲取信息，這些節點都是內核提供的統計信息。
sar -o temp.bin 1 600--------------------------------------------------sar將采樣數據保存在temp.bin中。
LC_ALL=C sar -A -f temp.bin > sar.txt----------------------------將保存的采樣數據temp.bin，轉換成更可讀性強的文本文件。
所以ksar的每一張圖標，都對應了內核統計信息。
sar是數據搬運整理工具，ksar是圖形化工具。
下面對每張圖標從ksar，到sar/sadc，最終到內核中每個數據。
2.2 ksar操作
2.2.1 數據導入
通過Data->Append from a file...從txt中加載數據，還有其他兩種數據來源方式。
2.2.2 數據導出
如果要將圖表導出，可以在每張圖標下面選擇Export PNG之類。
或者通過Export->Export to PDF...，選擇指定選項。
3. ksar解析
下面結合ksar圖表來逐項解析。
3.1 CPU信息
cpu_act讀取/proc/stat節點，解析其中cpu信息，包括合計cpu信息以及單個cpu信息。
/proc/stat中的信息包括一個合計以及多個cpu單獨統計信息。
從/proc/stat中讀取的信息都是從啟動以來的累計時間，在圖標中顯示的是一個時間段的差值。
然后計算不同模塊耗時占比。
void read_stat_cpu(struct stats_cpu *st_cpu, int nbr, unsigned long long *uptime, unsigned long long *uptime0) { FILE *fp; struct stats_cpu *st_cpu_i; struct stats_cpu sc; char line[8192]; int proc_nb; if ((fp = fopen(STAT, "r")) == NULL) {----------------------------------------------------/proc/stat fprintf(stderr, _("Cannot open %s: %s\n"), STAT, strerror(errno)); exit(2); } while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "cpu ", 4)) {------------------------------------------------------統計總cpu不同類別耗時，詳細信息參考/proc/stat。這里的信息和內核中一一對應。 memset(st_cpu, 0, STATS_CPU_SIZE); sscanf(line + 5, "%llu %llu %llu %llu %llu %llu %llu %llu %llu %llu", &st_cpu->cpu_user, &st_cpu->cpu_nice, &st_cpu->cpu_sys, &st_cpu->cpu_idle, &st_cpu->cpu_iowait, &st_cpu->cpu_hardirq, &st_cpu->cpu_softirq, &st_cpu->cpu_steal, &st_cpu->cpu_guest, &st_cpu->cpu_guest_nice); *uptime = st_cpu->cpu_user + st_cpu->cpu_nice + st_cpu->cpu_sys + st_cpu->cpu_idle + st_cpu->cpu_iowait + st_cpu->cpu_hardirq + st_cpu->cpu_steal + st_cpu->cpu_softirq; } else if (!strncmp(line, "cpu", 3)) { ... } fclose(fp); }
3.2 進程創建及切換
pcsw_act讀取/proc/stat節點，解析其中的ctxt和processes信息。
同樣的/proc/stat中，統計信息是啟動以來的累計值，圖標中現實的不同時間段的差值。
proc/s表示每秒創建的進程數目，cswch/s表示每秒進程切換次數。
void read_stat_pcsw(struct stats_pcsw *st_pcsw) { FILE *fp; char line[8192]; if ((fp = fopen(STAT, "r")) == NULL) return; while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "ctxt ", 5)) {--------------------------------------------ctxt是所有CPU的進程切換次數。 /* Read number of context switches */ sscanf(line + 5, "%llu", &st_pcsw->context_switch); } else if (!strncmp(line, "processes ", 10)) {---------------------------------是整個系統創建進程的次數，total_forks， /* Read number of processes created since system boot */ sscanf(line + 10, "%lu", &st_pcsw->processes); } } fclose(fp); }
3.3 swap信息
swap_act讀取/proc/vmstat節點，解析其中的pswpin和pswpout兩個信息。
pswpin/s表示系統每秒從swap分區讀入的頁面數量，即移除掉swap；pswpout/s表示系統每秒寫到swap分區的頁面數量，即產生swap。
void read_vmstat_swap(struct stats_swap *st_swap) { FILE *fp; char line[128]; if ((fp = fopen(VMSTAT, "r")) == NULL) return; while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "pswpin ", 7)) {-------------------------------對應PSWPIN，表示從swap分區讀入頁面。 /* Read number of swap pages brought in */ sscanf(line + 7, "%lu", &st_swap->pswpin); } else if (!strncmp(line, "pswpout ", 8)) {-------------------------對應PSWPOUT，表示將page寫入到swap分區。 /* Read number of swap pages brought out */ sscanf(line + 8, "%lu", &st_swap->pswpout); } } fclose(fp); }
3.4 頁面統計信息
paging_act統計/proc/vmstat中頁面交換統計信息。
Paging包括四張圖標，分別統計/proc/vmstat中的PGPGIN、PGPGOUT、PGFAULT、PGMAJFAULT等等信息。
pgpgin/s表示系統每秒從磁盤中paged多少KB；pgpgout/s則表示到磁盤中多少KB。
fault/s表示系統每秒產生的頁面異常數目，包括major和minor；majflt/s則表示頁面異常是從磁盤中產生的。
pgfree/s表示每秒放入到free list的頁面數量；pgscank/s、pgscand/s都表示掃描的頁面數量，只不過前者表示kswapd掃描結果，后者表示直接掃描；pgsteal/s都表示從pagecache和swapcache中回收的頁面數量。
%vmeff是pgsteal/pgscan表示頁面回收的效率。
看fault/s和pgfree/s存在一定關系，前者表示內存申請，后者表示內存釋放，存在先后關系。但是單位不一致，前者是次數，后者是頁面。
void read_vmstat_paging(struct stats_paging *st_paging) { FILE *fp; char line[128]; unsigned long pgtmp; if ((fp = fopen(VMSTAT, "r")) == NULL) return; st_paging->pgsteal = 0; st_paging->pgscan_kswapd = st_paging->pgscan_direct = 0; while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "pgpgin ", 7)) {------------------------------------------------pgpgin和pgpgout分別對應PGPGIN和PGPGOUT，在submit_io()中更新。指內存和塊設備志堅page數目，這里的page指的是disk sector。pgpgin表示從塊設備讀入，pgpgout寫入到跨設備。 /* Read number of pages the system paged in */ sscanf(line + 7, "%lu", &st_paging->pgpgin); } else if (!strncmp(line, "pgpgout ", 8)) { /* Read number of pages the system paged out */ sscanf(line + 8, "%lu", &st_paging->pgpgout); } else if (!strncmp(line, "pgfault ", 8)) {-------------------------------------------在vm_event_item[]中對應PGFAULT，在handle_mm_fault()中更新。統計產生page fault的信息。 /* Read number of faults (major+minor) made by the system */ sscanf(line + 8, "%lu", &st_paging->pgfault); } else if (!strncmp(line, "pgmajfault ", 11)) {---------------------------------------對應PGMAJFAULT，表示從磁盤而不是從內存中獲取數據。 /* Read number of faults (major only) made by the system */ sscanf(line + 11, "%lu", &st_paging->pgmajfault); } else if (!strncmp(line, "pgfree ", 7)) {--------------------------------------------對應PGFREE，統計釋放的頁面次數。 /* Read number of pages freed by the system */ sscanf(line + 7, "%lu", &st_paging->pgfree); } else if (!strncmp(line, "pgsteal_", 8)) {-------------------------------------------統計PGSTEAL_KSWAPD和PGSTEAL_DIRECT信息，表示系統回收的kswapd和直接回收的頁面數目。 /* Read number of pages stolen by the system */ sscanf(strchr(line, ' '), "%lu", &pgtmp); st_paging->pgsteal += pgtmp; } else if (!strncmp(line, "pgscan_kswapd", 13)) {--------------------------------------統計PGSCAN_KSWAPD信息，表示從系統啟動到現在kswapd后台進程掃描的頁面數。 /* Read number of pages scanned by the kswapd daemon */ sscanf(strchr(line, ' '), "%lu", &pgtmp); st_paging->pgscan_kswapd += pgtmp; } else if (!strncmp(line, "pgscan_direct", 13)) {--------------------------------------統計PGSCAN_DIRECT和PGSCAN_DIRECT_THROTTLE，統計世界回收頁面數。 /* Read number of pages scanned directly */ sscanf(strchr(line, ' '), "%lu", &pgtmp); st_paging->pgscan_direct += pgtmp; } } fclose(fp); }
3.5 IO統計信息
io_act從/proc/diskstats讀取IO統計信息。
上面的tps、rtps、wtps、bread/s、bwrtn/s分別對應采樣結果dk_drive、dk_drive_rio、dk_drive_wio、dk_drive_rblk、dk_drive_wblk。
其中tps為rtps和wtps之和，表示完成讀寫次數；bread/s和bwrtn/s表示讀、寫扇區數目。
這5個參數都來源於diskstats_show()函數。
void read_diskstats_io(struct stats_io *st_io) { FILE *fp; char line[256]; char dev_name[MAX_NAME_LEN]; unsigned int major, minor; unsigned long rd_ios, wr_ios, rd_sec, wr_sec; if ((fp = fopen(DISKSTATS, "r")) == NULL) return; while (fgets(line, sizeof(line), fp) != NULL) { if (sscanf(line, "%u %u %s %lu %*u %lu %*u %lu %*u %lu",-----------scanf的%u之間的星號，表示跳過此輸入。 &major, &minor, dev_name, &rd_ios, &rd_sec, &wr_ios, &wr_sec) == 7) { if (is_device(dev_name, IGNORE_VIRTUAL_DEVICES)) {--------------虛擬設備沒有/sys/block/<device>/device，此特性用以判斷dev_name對應的設備是否是真實的塊設備。 st_io->dk_drive += (unsigned long long) rd_ios + (unsigned long long) wr_ios; st_io->dk_drive_rio += rd_ios; st_io->dk_drive_rblk += rd_sec; st_io->dk_drive_wio += wr_ios; st_io->dk_drive_wblk += wr_sec; } } } fclose(fp); }
/proc/diskstats來源於內核的diskstats_show()
static int diskstats_show(struct seq_file *seqf, void *v) { struct gendisk *gp = v; struct disk_part_iter piter; struct hd_struct *hd; char buf[BDEVNAME_SIZE]; int cpu; /* if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) seq_puts(seqf, "major minor name" " rio rmerge rsect ruse wio wmerge " "wsect wuse running use aveq" "\n\n"); */ disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0); while ((hd = disk_part_iter_next(&piter))) { cpu = part_stat_lock(); part_round_stats(cpu, hd); part_stat_unlock(); seq_printf(seqf, "%4d %7d %s %lu %lu %lu " "%u %lu %lu %lu %u %u %u %u\n", MAJOR(part_devt(hd)), MINOR(part_devt(hd)), disk_name(gp, hd->partno, buf), part_stat_read(hd, ios[READ]),-------------------------------------------------成功完成讀的次數。 part_stat_read(hd, merges[READ]),----------------------------------------------合並讀次數，為了效率可能會合並相鄰的讀和寫。 part_stat_read(hd, sectors[READ]),---------------------------------------------讀扇區的次數。 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),-----------------------------讀花的時間，這里是所有讀操作所花費的毫秒數。 part_stat_read(hd, ios[WRITE]),------------------------------------------------成功完成寫的次數。 part_stat_read(hd, merges[WRITE]),---------------------------------------------合並寫次數。 part_stat_read(hd, sectors[WRITE]),--------------------------------------------寫扇區次數。 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),----------------------------寫花的時間。 part_in_flight(hd), jiffies_to_msecs(part_stat_read(hd, io_ticks)), jiffies_to_msecs(part_stat_read(hd, time_in_queue)) ); } disk_part_iter_exit(&piter); return 0; }
3.6 內存及swap統計信息
memory_act統計/proc/meminfo的內存和swap信息。
kbmemfree是所有free內存大小，kbavail是扣除保留內存，加上部分pagecache和可回收內存；kbmemused是目前系統內存使用量，不包括內核使用量。
%memused是內存使用占總內存百分比。
kbcommit是當前場景系統需要使用到的內存量，實際上可能並沒有申請這么多內存。這是因為分配的內存只有在使用到時，才會產生缺頁異常。
這里的kbcommit和頁面統計信息相呼應，此處kbcommit突然增加，產生了很多page fault。
從圖表中的名稱大概就能找到對應的/proc/meminfo中的字符項。
其中%memused和%commit查看stub_print_memory_stats()，%memused為(MemTotal-MemFree)/MemTotal。
%commit為Commited_AS/(MemTotal+SwapTotal)。
在/proc/meminfo中看到的CommitLimit可能是proc/sys/vm/overcommit_kbytes；或者按照當前可以用內存乘以overcommit_ratio這個比例得到。
在實際使用中，還需要結合overcommit_memory類型來看，參考overcommit_memory和overcommit_ratio。
unsigned long vm_commit_limit(void) { unsigned long allowed; if (sysctl_overcommit_kbytes) allowed = sysctl_overcommit_kbytes >> (PAGE_SHIFT - 10); else allowed = ((totalram_pages - hugetlb_total_pages()) * sysctl_overcommit_ratio / 100); allowed += total_swap_pages; return allowed; }
read_meminfo()從/proc/meminfo中解析數據。
void read_meminfo(struct stats_memory *st_memory) { FILE *fp; char line[128]; if ((fp = fopen(MEMINFO, "r")) == NULL) return; while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "MemTotal:", 9)) { /* Read the total amount of memory in kB */ sscanf(line + 9, "%lu", &st_memory->tlmkb); } else if (!strncmp(line, "MemFree:", 8)) { /* Read the amount of free memory in kB */ sscanf(line + 8, "%lu", &st_memory->frmkb); } else if (!strncmp(line, "MemAvailable:", 13)) { /* Read the amount of available memory in kB */ sscanf(line + 13, "%lu", &st_memory->availablekb); } else if (!strncmp(line, "Buffers:", 8)) { /* Read the amount of buffered memory in kB */ sscanf(line + 8, "%lu", &st_memory->bufkb); } else if (!strncmp(line, "Cached:", 7)) { /* Read the amount of cached memory in kB */ sscanf(line + 7, "%lu", &st_memory->camkb); } else if (!strncmp(line, "SwapCached:", 11)) { /* Read the amount of cached swap in kB */ sscanf(line + 11, "%lu", &st_memory->caskb); } else if (!strncmp(line, "Active:", 7)) { /* Read the amount of active memory in kB */ sscanf(line + 7, "%lu", &st_memory->activekb); } else if (!strncmp(line, "Inactive:", 9)) { /* Read the amount of inactive memory in kB */ sscanf(line + 9, "%lu", &st_memory->inactkb); } else if (!strncmp(line, "SwapTotal:", 10)) { /* Read the total amount of swap memory in kB */ sscanf(line + 10, "%lu", &st_memory->tlskb); } else if (!strncmp(line, "SwapFree:", 9)) { /* Read the amount of free swap memory in kB */ sscanf(line + 9, "%lu", &st_memory->frskb); } else if (!strncmp(line, "Dirty:", 6)) { /* Read the amount of dirty memory in kB */ sscanf(line + 6, "%lu", &st_memory->dirtykb); } else if (!strncmp(line, "Committed_AS:", 13)) { /* Read the amount of commited memory in kB */ sscanf(line + 13, "%lu", &st_memory->comkb); } else if (!strncmp(line, "AnonPages:", 10)) { /* Read the amount of pages mapped into userspace page tables in kB */ sscanf(line + 10, "%lu", &st_memory->anonpgkb); } else if (!strncmp(line, "Slab:", 5)) { /* Read the amount of in-kernel data structures cache in kB */ sscanf(line + 5, "%lu", &st_memory->slabkb); } else if (!strncmp(line, "KernelStack:", 12)) { /* Read the kernel stack utilization in kB */ sscanf(line + 12, "%lu", &st_memory->kstackkb); } else if (!strncmp(line, "PageTables:", 11)) { /* Read the amount of memory dedicated to the lowest level of page tables in kB */ sscanf(line + 11, "%lu", &st_memory->pgtblkb); } else if (!strncmp(line, "VmallocUsed:", 12)) { /* Read the amount of vmalloc area which is used in kB */ sscanf(line + 12, "%lu", &st_memory->vmusedkb); } } fclose(fp); }
下圖的kbswpfree、kbswpused、kbswpcad通過字面意思即可知其對應的meminfo項為SwapFree、SwapTotal-SwapFree、SwapCached。
3.7 中斷統計信息
中斷你統計信息同樣來自/proc/stat，解析intr字段。第一個是啟動以來所有中斷觸發次數，后面是單個中斷觸發次數。
生成的表格中包含了對應的圖標sum，以及每個中斷統計信息。
void read_stat_irq(struct stats_irq *st_irq, int nbr) { FILE *fp; struct stats_irq *st_irq_i; char line[8192]; int i, pos; if ((fp = fopen(STAT, "r")) == NULL) return; while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "intr ", 5)) {-------------------------------------------------/proc/stat的intr字段。 /* Read total number of interrupts received since system boot */ sscanf(line + 5, "%llu", &st_irq->irq_nr); pos = strcspn(line + 5, " ") + 5; for (i = 1; i < nbr; i++) { st_irq_i = st_irq + i; sscanf(line + pos, " %llu", &st_irq_i->irq_nr); pos += strcspn(line + pos + 1, " ") + 1; } } } fclose(fp); }
3.8
ktables_act
#define FDENTRY_STATE "/proc/sys/fs/dentry-state" #define FFILE_NR "/proc/sys/fs/file-nr" #define FINODE_STATE "/proc/sys/fs/inode-state" #define PTY_NR "/proc/sys/kernel/pty/nr" void read_kernel_tables(struct stats_ktables *st_ktables) { FILE *fp; unsigned int parm; int rc = 0; /* Open /proc/sys/fs/dentry-state file */ if ((fp = fopen(FDENTRY_STATE, "r")) != NULL) { rc = fscanf(fp, "%*d %u", &st_ktables->dentry_stat); fclose(fp); if (rc == 0) { st_ktables->dentry_stat = 0; } } /* Open /proc/sys/fs/file-nr file */ if ((fp = fopen(FFILE_NR, "r")) != NULL) { rc = fscanf(fp, "%u %u", &st_ktables->file_used, &parm); fclose(fp); /* * The number of used handles is the number of allocated ones * minus the number of free ones. */ if (rc == 2) { st_ktables->file_used -= parm; } else { st_ktables->file_used = 0; } } /* Open /proc/sys/fs/inode-state file */ if ((fp = fopen(FINODE_STATE, "r")) != NULL) { rc = fscanf(fp, "%u %u", &st_ktables->inode_used, &parm); fclose(fp); /* * The number of inuse inodes is the number of allocated ones * minus the number of free ones. */ if (rc == 2) { st_ktables->inode_used -= parm; } else { st_ktables->inode_used = 0; } } /* Open /proc/sys/kernel/pty/nr file */ if ((fp = fopen(PTY_NR, "r")) != NULL) { rc = fscanf(fp, "%u", &st_ktables->pty_nr); fclose(fp); if (rc == 0) { st_ktables->pty_nr = 0; } } }
3.9 load和queue統計信息
queue_act從/proc/loadavg獲取load信息，從/proc/stat獲取queue統計信息。
ldavg-1、ldavg-5、ldavg-15分別表示1、5、15分鍾內進程對立中平均進程數目，包括正在運行的進程和准備好等待運行的進程數目。
runq-sz表示運行隊列大小，plist-sz表示所有進程總數nr_threads。
runq-sz大表示等待運行的進程數目較多。
blocked表示處於iowait狀態的進程數目。
void read_loadavg(struct stats_queue *st_queue) { FILE *fp; char line[8192]; int load_tmp[3]; int rc; if ((fp = fopen(LOADAVG, "r")) == NULL) return; /* Read load averages and queue length */ rc = fscanf(fp, "%d.%u %d.%u %d.%u %lu/%u %*d\n",------------------------------可以看出從/proc/loadavg中獲取信息，然后最后一個數據忽略。 &load_tmp[0], &st_queue->load_avg_1, &load_tmp[1], &st_queue->load_avg_5, &load_tmp[2], &st_queue->load_avg_15, &st_queue->nr_running, &st_queue->nr_threads); fclose(fp); if (rc < 8) return; st_queue->load_avg_1 += load_tmp[0] * 100; st_queue->load_avg_5 += load_tmp[1] * 100; st_queue->load_avg_15 += load_tmp[2] * 100; if (st_queue->nr_running) { /* Do not take current process into account */ st_queue->nr_running--; } /* Read nr of tasks blocked from /proc/stat */ if ((fp = fopen(STAT, "r")) == NULL) return; while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "procs_blocked ", 14)) {--------------------------------從/proc/stat中獲取procs_blocked項，表示處於iowait狀態的進程數目。 /* Read number of processes blocked */ sscanf(line + 14, "%lu", &st_queue->procs_blocked); break; } } fclose(fp); }
3.10 塊設備統計信息
disk_act和io_act一樣是從/proc/diskstats中獲取統計信息，
圖標中的四項tps表示塊設備讀寫次數頻率；rkB/s和wkB/s表示塊設備讀寫速率；await表示一次讀或寫耗時。
tps值大說明此段時間讀寫較頻繁。
await值大表示當前讀寫耗時較大，可能存在問題。
__print_funct_t print_disk_stats(struct activity *a, int prev, int curr, unsigned long long itv) { ... for (i = 0; i < a->nr; i++) { ... printf("%-11s", timestamp[curr]); cprintf_in(IS_STR, " %9s", dev_name, 0);-----------------------------sdc是設備當前統計信息，sdp是設備前一次統計信息，itv是兩次時間間隔。 cprintf_f(NO_UNIT, 1, 9, 2, S_VALUE(sdp->nr_ios, sdc->nr_ios, itv));----------------------對應圖標中的tps，nr_ios表示塊設備讀寫次數之和。tps表示塊設備讀寫頻率。 cprintf_f(unit, 2, 9, 2, S_VALUE(sdp->rd_sect, sdc->rd_sect, itv) / 2, S_VALUE(sdp->wr_sect, sdc->wr_sect, itv) / 2);-----------------對應圖標的rkB/s和wkB/s，同樣rd_sect和wr_sect可以計算出以KB為單位的讀寫速率。 /* See iostat for explanations */ cprintf_f(unit, 1, 9, 2, xds.arqsz / 2); cprintf_f(NO_UNIT, 3, 9, 2, S_VALUE(sdp->rq_ticks, sdc->rq_ticks, itv) / 1000.0, xds.await,------------------------------------------------------對應await，在compute_ext_disk_stats()中計算。表示每次讀寫平均耗時。 xds.svctm); cprintf_pc(DISPLAY_UNIT(flags), 1, 9, 2, xds.util / 10.0); printf("\n"); } }
3.11 CPU頻率
pwr_cpufreq_act從/proc/cpuinfo中獲取頻率信息，CPU Frequency all顯示的是所有CPU頻率的平均值。
從下面代碼分析可知，從關鍵詞processor中獲取CPU號，從cpu MHz獲取CPU頻率。
void read_cpuinfo(struct stats_pwr_cpufreq *st_pwr_cpufreq, int nbr) { FILE *fp; struct stats_pwr_cpufreq *st_pwr_cpufreq_i; char line[1024]; int nr = 0; unsigned int proc_nb = 0, ifreq, dfreq; if ((fp = fopen(CPUINFO, "r")) == NULL) return; st_pwr_cpufreq->cpufreq = 0; while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "processor\t", 10)) { sscanf(strchr(line, ':') + 1, "%u", &proc_nb); } /* Entry in /proc/cpuinfo is different between Intel and Power architectures */ else if (!strncmp(line, "cpu MHz\t", 8) || !strncmp(line, "clock\t", 6)) { sscanf(strchr(line, ':') + 1, "%u.%u", &ifreq, &dfreq); if (proc_nb < (nbr - 1)) { st_pwr_cpufreq_i = st_pwr_cpufreq + proc_nb + 1; st_pwr_cpufreq_i->cpufreq = ifreq * 100 + dfreq / 10; st_pwr_cpufreq->cpufreq += st_pwr_cpufreq_i->cpufreq; nr++; } else if (!proc_nb && (nbr == 1)) { st_pwr_cpufreq->cpufreq = ifreq * 100 + dfreq / 10; } } } fclose(fp); ... }
3.12 統計串口線信息
serial_act從/proc/tty/driver/serial獲取串口線統計信息。
void read_tty_driver_serial(struct stats_serial *st_serial, int nbr) { FILE *fp; struct stats_serial *st_serial_i; int sl = 0; char line[256]; char *p; if ((fp = fopen(SERIAL, "r")) == NULL) return; while ((fgets(line, sizeof(line), fp) != NULL) && (sl < nbr)) { if ((p = strstr(line, "tx:")) != NULL) { st_serial_i = st_serial + sl; sscanf(line, "%u", &st_serial_i->line); /* * A value of 0 means an unused structure. * So increment it to make sure it is not zero. */ (st_serial_i->line)++; /* * Read the number of chars transmitted and received by * current serial line. */ sscanf(p + 3, "%u", &st_serial_i->tx); if ((p = strstr(line, "rx:")) != NULL) { sscanf(p + 3, "%u", &st_serial_i->rx); } if ((p = strstr(line, "fe:")) != NULL) { sscanf(p + 3, "%u", &st_serial_i->frame); } if ((p = strstr(line, "pe:")) != NULL) { sscanf(p + 3, "%u", &st_serial_i->parity); } if ((p = strstr(line, "brk:")) != NULL) { sscanf(p + 4, "%u", &st_serial_i->brk); } if ((p = strstr(line, "oe:")) != NULL) { sscanf(p + 3, "%u", &st_serial_i->overrun); } sl++; } } fclose(fp); }
3.13 獲取網口統計信息
net_dev_act從/proc/net/dev中獲取網口的統計信息，從/sys/class/net/xxx/duplex和/sys/class/net/xxx/speed中獲取雙工和速度信息。
從下面關於/proc/net/dev解釋可知，rxpck對應rx_packets，rxpck/s就是每秒接收packet數；txpck/s就是每個發送packet數目。
rxkB/s是每秒接受多少k字節；txkB/s是每秒發送多少k字節。
rxmcst/s表示每秒接收到的multicast packet數目。
%ifutil跟每秒發送接收的KB數有關，還跟全雙工，單雙工有關，以及設備的速度有關；表示網絡設備使用率。
雙工信息從/sys/class/net/%s/duplex獲取，速度信息從從/sys/class/net/%s/speed獲取。
從此圖看，網絡的使用率並不高，說明網絡不頻繁。
int read_net_dev(struct stats_net_dev *st_net_dev, int nbr) { FILE *fp; struct stats_net_dev *st_net_dev_i; char line[256]; char iface[MAX_IFACE_LEN]; int dev = 0; int pos; if ((fp = fopen(NET_DEV, "r")) == NULL)----------------------------------------------------------------從/proc/net/dev獲取信息。 return 0; while ((fgets(line, sizeof(line), fp) != NULL) && (dev < nbr)) { pos = strcspn(line, ":"); if (pos < strlen(line)) { st_net_dev_i = st_net_dev + dev; strncpy(iface, line, MINIMUM(pos, MAX_IFACE_LEN - 1)); iface[MINIMUM(pos, MAX_IFACE_LEN - 1)] = '\0'; sscanf(iface, "%s", st_net_dev_i->interface); /* Skip heading spaces */ sscanf(line + pos + 1, "%llu %llu %*u %*u %*u %*u %llu %llu %llu %llu " "%*u %*u %*u %*u %*u %llu", &st_net_dev_i->rx_bytes, &st_net_dev_i->rx_packets, &st_net_dev_i->rx_compressed, &st_net_dev_i->multicast, &st_net_dev_i->tx_bytes, &st_net_dev_i->tx_packets, &st_net_dev_i->tx_compressed); dev++; } } fclose(fp); return dev; } void read_if_info(struct stats_net_dev *st_net_dev, int nbr) { FILE *fp; struct stats_net_dev *st_net_dev_i; char filename[128], duplex[32]; int dev, n; for (dev = 0; dev < nbr; dev++) { st_net_dev_i = st_net_dev + dev; /* Read speed info */ sprintf(filename, IF_DUPLEX, st_net_dev_i->interface);----------------------------------------------從/sys/class/net/%s/duplex獲取信息。 if ((fp = fopen(filename, "r")) == NULL) /* Cannot read NIC duplex */ continue; n = fscanf(fp, "%31s", duplex); fclose(fp); if (n != 1) /* Cannot read NIC duplex */ continue; if (!strcmp(duplex, K_DUPLEX_FULL)) { st_net_dev_i->duplex = C_DUPLEX_FULL; } else if (!strcmp(duplex, K_DUPLEX_HALF)) { st_net_dev_i->duplex = C_DUPLEX_HALF; } else continue; /* Read speed info */ sprintf(filename, IF_SPEED, st_net_dev_i->interface);--------------------------------------------------從/sys/class/net/%s/speed獲取信息。 if ((fp = fopen(filename, "r")) == NULL) /* Cannot read NIC speed */ continue; n = fscanf(fp, "%u", &st_net_dev_i->speed); fclose(fp); if (n != 1) { st_net_dev_i->speed = 0; } } }
3.13.1 /proc/net/dev
/proc/net/dev節點在dev_proc_net_init()中創建，對應的fops是dev_seq_fops。
這里重點看一下dev_seq_show()：
static int dev_seq_show(struct seq_file *seq, void *v) { if (v == SEQ_START_TOKEN) seq_puts(seq, "Inter-| Receive " " | Transmit\n" " face |bytes packets errs drop fifo frame " "compressed multicast|bytes packets errs " "drop fifo colls carrier compressed\n"); else dev_seq_printf_stats(seq, v); return 0; } static void dev_seq_printf_stats(struct seq_file *seq, struct net_device *dev) { struct rtnl_link_stats64 temp; const struct rtnl_link_stats64 *stats = dev_get_stats(dev, &temp); seq_printf(seq, "%6s: %7llu %7llu %4llu %4llu %4llu %5llu %10llu %9llu " "%8llu %7llu %4llu %4llu %4llu %5llu %7llu %10llu\n", dev->name, stats->rx_bytes, stats->rx_packets, stats->rx_errors, stats->rx_dropped + stats->rx_missed_errors, stats->rx_fifo_errors, stats->rx_length_errors + stats->rx_over_errors + stats->rx_crc_errors + stats->rx_frame_errors, stats->rx_compressed, stats->multicast, stats->tx_bytes, stats->tx_packets, stats->tx_errors, stats->tx_dropped, stats->tx_fifo_errors, stats->collisions, stats->tx_carrier_errors + stats->tx_aborted_errors + stats->tx_window_errors + stats->tx_heartbeat_errors, stats->tx_compressed); }
3.14 網絡錯誤統計信息
net_edev_act從/proc/net/dev網絡錯誤統計信息。
網絡錯誤信息數據來源和網絡信息一樣，都來自於/proc/net/dev。
關鍵結構體也是struct rtnl_link_stats64。
void read_net_edev(struct stats_net_edev *st_net_edev, int nbr) { FILE *fp; struct stats_net_edev *st_net_edev_i; static char line[256]; char iface[MAX_IFACE_LEN]; int dev = 0; int pos; if ((fp = fopen(NET_DEV, "r")) == NULL) return; while ((fgets(line, sizeof(line), fp) != NULL) && (dev < nbr)) { pos = strcspn(line, ":"); if (pos < strlen(line)) { st_net_edev_i = st_net_edev + dev; strncpy(iface, line, MINIMUM(pos, MAX_IFACE_LEN - 1)); iface[MINIMUM(pos, MAX_IFACE_LEN - 1)] = '\0'; sscanf(iface, "%s", st_net_edev_i->interface); /* Skip heading spaces */ sscanf(line + pos + 1, "%*u %*u %llu %llu %llu %llu %*u %*u %*u %*u " "%llu %llu %llu %llu %llu", &st_net_edev_i->rx_errors,------------------對應rx_erros，表示接受到的bad packet。 &st_net_edev_i->rx_dropped,-----------------對應rx_dropped+rx_missed_errors。 &st_net_edev_i->rx_fifo_errors,-------------對應rx_fifo_errors。 &st_net_edev_i->rx_frame_errors,------------對應rx_length_errors+rx_over_errros。 &st_net_edev_i->tx_errors, &st_net_edev_i->tx_dropped, &st_net_edev_i->tx_fifo_errors, &st_net_edev_i->collisions,-----------------對應collisions &st_net_edev_i->tx_carrier_errors);---------包括tx_carrier_errors+tx_aborted_errors+tx_window_errors+tx_heartbeat_errors。
dev++; } } fclose(fp); }
獲取網絡設備統計信息的核心數據結構式struct rtnl_link_stat64。
struct rtnl_link_stats64 { __u64 rx_packets; /* total packets received */ __u64 tx_packets; /* total packets transmitted */ __u64 rx_bytes; /* total bytes received */ __u64 tx_bytes; /* total bytes transmitted */ __u64 rx_errors; /* bad packets received */ __u64 tx_errors; /* packet transmit problems */ __u64 rx_dropped; /* no space in linux buffers */ __u64 tx_dropped; /* no space available in linux */ __u64 multicast; /* multicast packets received */ __u64 collisions; /* detailed rx_errors: */ __u64 rx_length_errors; __u64 rx_over_errors; /* receiver ring buff overflow */ __u64 rx_crc_errors; /* recved pkt with crc error */ __u64 rx_frame_errors; /* recv'd frame alignment error */ __u64 rx_fifo_errors; /* recv'r fifo overrun */ __u64 rx_missed_errors; /* receiver missed packet */ /* detailed tx_errors */ __u64 tx_aborted_errors; __u64 tx_carrier_errors; __u64 tx_fifo_errors; __u64 tx_heartbeat_errors; __u64 tx_window_errors; /* for cslip etc */ __u64 rx_compressed; __u64 tx_compressed; __u64 rx_nohandler; /* dropped, no handler found */ };
3.15 NFS客戶端統計信息
net_nfs_act讀取/proc/net/rpc/nfs來分析作為nfs客戶端的統計信息。
void read_net_nfs(struct stats_net_nfs *st_net_nfs) { FILE *fp; char line[256]; unsigned int getattcnt = 0, accesscnt = 0, readcnt = 0, writecnt = 0; if ((fp = fopen(NET_RPC_NFS, "r")) == NULL) return; memset(st_net_nfs, 0, STATS_NET_NFS_SIZE); while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "rpc ", 4)) { sscanf(line + 4, "%u %u", &st_net_nfs->nfs_rpccnt, &st_net_nfs->nfs_rpcretrans); } else if (!strncmp(line, "proc3 ", 6)) { sscanf(line + 6, "%*u %*u %u %*u %*u %u %*u %u %u", &getattcnt, &accesscnt, &readcnt, &writecnt); st_net_nfs->nfs_getattcnt += getattcnt; st_net_nfs->nfs_accesscnt += accesscnt; st_net_nfs->nfs_readcnt += readcnt; st_net_nfs->nfs_writecnt += writecnt; } else if (!strncmp(line, "proc4 ", 6)) { sscanf(line + 6, "%*u %*u %u %u " "%*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %u %u", &readcnt, &writecnt, &accesscnt, &getattcnt); st_net_nfs->nfs_getattcnt += getattcnt; st_net_nfs->nfs_accesscnt += accesscnt; st_net_nfs->nfs_readcnt += readcnt; st_net_nfs->nfs_writecnt += writecnt; } } fclose(fp); }
3.16 NFS服務端統計信息
net_nfsd_act
#define NET_RPC_NFSD "/proc/net/rpc/nfsd"
void read_net_nfsd(struct stats_net_nfsd *st_net_nfsd) { FILE *fp; char line[256]; unsigned int getattcnt = 0, accesscnt = 0, readcnt = 0, writecnt = 0; if ((fp = fopen(NET_RPC_NFSD, "r")) == NULL) return; memset(st_net_nfsd, 0, STATS_NET_NFSD_SIZE); while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "rc ", 3)) { sscanf(line + 3, "%u %u", &st_net_nfsd->nfsd_rchits, &st_net_nfsd->nfsd_rcmisses); } else if (!strncmp(line, "net ", 4)) { sscanf(line + 4, "%u %u %u", &st_net_nfsd->nfsd_netcnt, &st_net_nfsd->nfsd_netudpcnt, &st_net_nfsd->nfsd_nettcpcnt); } else if (!strncmp(line, "rpc ", 4)) { sscanf(line + 4, "%u %u", &st_net_nfsd->nfsd_rpccnt, &st_net_nfsd->nfsd_rpcbad); } else if (!strncmp(line, "proc3 ", 6)) { sscanf(line + 6, "%*u %*u %u %*u %*u %u %*u %u %u", &getattcnt, &accesscnt, &readcnt, &writecnt); st_net_nfsd->nfsd_getattcnt += getattcnt; st_net_nfsd->nfsd_accesscnt += accesscnt; st_net_nfsd->nfsd_readcnt += readcnt; st_net_nfsd->nfsd_writecnt += writecnt; } else if (!strncmp(line, "proc4ops ", 9)) { sscanf(line + 9, "%*u %*u %*u %*u %u " "%*u %*u %*u %*u %*u %u " "%*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %u " "%*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %u", &accesscnt, &getattcnt, &readcnt, &writecnt); st_net_nfsd->nfsd_getattcnt += getattcnt; st_net_nfsd->nfsd_accesscnt += accesscnt; st_net_nfsd->nfsd_readcnt += readcnt; st_net_nfsd->nfsd_writecnt += writecnt; } } fclose(fp); }
3.17 網絡socket統計信息
net_sock_act讀取/proc/net/sockstat統計TCP和UDP信息。
/proc/net/sockstat的數據來源於sockstat_seq_show()，這里面統計了當前系統socket的使用情況。
static int sockstat_seq_show(struct seq_file *seq, void *v) { struct net *net = seq->private; unsigned int frag_mem; int orphans, sockets; local_bh_disable(); orphans = percpu_counter_sum_positive(&tcp_orphan_count); sockets = proto_sockets_allocated_sum_positive(&tcp_prot); local_bh_enable(); socket_seq_show(seq);---------------------------------------------------------從sockets_ in_use中獲取每個CPU在使用中的socket數目。顯示內容sockets: used xxx。tw表示TimeWait的socket數目。 seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %ld\n", sock_prot_inuse_get(net, &tcp_prot), orphans, atomic_read(&tcp_death_row.tw_count), sockets, proto_memory_allocated(&tcp_prot));------------------------------------第一個參數表示TCP協議使用的socket數目。 seq_printf(seq, "UDP: inuse %d mem %ld\n", sock_prot_inuse_get(net, &udp_prot), proto_memory_allocated(&udp_prot));------------------------------------第一個參數表示UDP協議使用的socket數目。 seq_printf(seq, "UDPLITE: inuse %d\n", sock_prot_inuse_get(net, &udplite_prot)); seq_printf(seq, "RAW: inuse %d\n", sock_prot_inuse_get(net, &raw_prot));----------------------------------RAW類型socket使用數目。 frag_mem = ip_frag_mem(net); seq_printf(seq, "FRAG: inuse %u memory %u\n", !!frag_mem, frag_mem); return 0; }
Sockets圖表中的數據totsck、tcpsck、udpsck、rawsck、tcp-tw、ip-frag分別來源於read_net_sock()函數解析。
可見totsck表示系統所有Socket數目，其他子類表示不同類型socket數目。
void read_net_sock(struct stats_net_sock *st_net_sock) { FILE *fp; char line[96]; char *p; if ((fp = fopen(NET_SOCKSTAT, "r")) == NULL) return; while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "sockets:", 8)) { /* Sockets */ sscanf(line + 14, "%u", &st_net_sock->sock_inuse); } else if (!strncmp(line, "TCP:", 4)) { /* TCP sockets */ sscanf(line + 11, "%u", &st_net_sock->tcp_inuse); if ((p = strstr(line, "tw")) != NULL) { sscanf(p + 2, "%u", &st_net_sock->tcp_tw); } } else if (!strncmp(line, "UDP:", 4)) { /* UDP sockets */ sscanf(line + 11, "%u", &st_net_sock->udp_inuse); } else if (!strncmp(line, "RAW:", 4)) { /* RAW sockets */ sscanf(line + 11, "%u", &st_net_sock->raw_inuse); } else if (!strncmp(line, "FRAG:", 5)) { /* FRAGments */ sscanf(line + 12, "%u", &st_net_sock->frag_inuse); } } fclose(fp); }
3.18 huge頁面統計信息
huge_act通過讀取/proc/meminfo獲取huge頁面統計信息。
void read_meminfo_huge(struct stats_huge *st_huge) { FILE *fp; char line[128]; unsigned long szhkb = 0; if ((fp = fopen(MEMINFO, "r")) == NULL) return; while (fgets(line, sizeof(line), fp) != NULL) { if (!strncmp(line, "HugePages_Total:", 16)) { /* Read the total number of huge pages */ sscanf(line + 16, "%lu", &st_huge->tlhkb); } else if (!strncmp(line, "HugePages_Free:", 15)) { /* Read the number of free huge pages */ sscanf(line + 15, "%lu", &st_huge->frhkb); } else if (!strncmp(line, "Hugepagesize:", 13)) { /* Read the default size of a huge page in kB */ sscanf(line + 13, "%lu", &szhkb); } } fclose(fp); /* We want huge pages stats in kB and not expressed in a number of pages */ st_huge->tlhkb *= szhkb; st_huge->frhkb *= szhkb; }
3.19 softnet統計信息
softnet_act通過/proc/net/softnet_stat獲取統計信息，
void read_softnet(struct stats_softnet *st_softnet, int nbr) { FILE *fp; struct stats_softnet *st_softnet_i; char line[1024]; unsigned int proc_nb = 1; /* Open /proc/net/softnet_stat file */ if ((fp = fopen(NET_SOFTNET, "r")) == NULL) return; /* * Init a structure that will contain the values for CPU "all". * CPU "all" doesn't exist in /proc/net/softnet_stat file, so * we compute its values as the sum of the values of each CPU. */ memset(st_softnet, 0, sizeof(struct stats_softnet)); while ((fgets(line, sizeof(line), fp) != NULL) && (proc_nb < nbr)) { st_softnet_i = st_softnet + proc_nb++; sscanf(line, "%x %x %x %*x %*x %*x %*x %*x %*x %x %x", &st_softnet_i->processed, &st_softnet_i->dropped, &st_softnet_i->time_squeeze, &st_softnet_i->received_rps, &st_softnet_i->flow_limit); st_softnet->processed += st_softnet_i->processed; st_softnet->dropped += st_softnet_i->dropped; st_softnet->time_squeeze += st_softnet_i->time_squeeze; st_softnet->received_rps += st_softnet_i->received_rps; st_softnet->flow_limit += st_softnet_i->flow_limit; } fclose(fp); }
/proc/net/softnet_stat數據來源於softnet_seq_show()。
static int softnet_seq_show(struct seq_file *seq, void *v) { struct softnet_data *sd = v; unsigned int flow_limit_count = 0; #ifdef CONFIG_NET_FLOW_LIMIT struct sd_flow_limit *fl; rcu_read_lock(); fl = rcu_dereference(sd->flow_limit); if (fl) flow_limit_count = fl->count; rcu_read_unlock(); #endif seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x\n", sd->processed, sd->dropped, sd->time_squeeze, 0, 0, 0, 0, 0, /* was fastroute */ 0, /* was cpu_collision */ sd->received_rps, flow_limit_count); return 0; }