Date: Tue, 2 Sep 2025 05:49:30 GMT From: Poul-Henning Kamp <phk@FreeBSD.org> To: src-committers@FreeBSD.org, dev-commits-src-all@FreeBSD.org, dev-commits-src-main@FreeBSD.org Subject: git: e88b7dcac870 - main - Recoverdisk: Refine the determination of defaults Message-ID: <202509020549.5825nUbh082126@gitrepo.freebsd.org>
next in thread | raw e-mail | index | archive | help
The branch main has been updated by phk: URL: https://cgit.FreeBSD.org/src/commit/?id=e88b7dcac8701b6d925b8c7c66eada5143c37e16 commit e88b7dcac8701b6d925b8c7c66eada5143c37e16 Author: Poul-Henning Kamp <phk@FreeBSD.org> AuthorDate: 2025-09-02 05:48:05 +0000 Commit: Poul-Henning Kamp <phk@FreeBSD.org> CommitDate: 2025-09-02 05:48:05 +0000 Recoverdisk: Refine the determination of defaults --- sbin/recoverdisk/recoverdisk.1 | 6 ++++ sbin/recoverdisk/recoverdisk.c | 72 +++++++++++++++++++++++++++++++----------- 2 files changed, 59 insertions(+), 19 deletions(-) diff --git a/sbin/recoverdisk/recoverdisk.1 b/sbin/recoverdisk/recoverdisk.1 index 9f1deb4c0c23..90849755ea0c 100644 --- a/sbin/recoverdisk/recoverdisk.1 +++ b/sbin/recoverdisk/recoverdisk.1 @@ -31,6 +31,7 @@ .Sh SYNOPSIS .Nm .Op Fl b Ar bigsize +.Op Fl i Ar interval .Op Fl r Ar readlist .Op Fl s Ar interval .Op Fl u Ar pattern @@ -109,6 +110,11 @@ reports for character and block devices or if .Ar source is a regular file. +.It Fl i Ar pause +.Xr sleep 3 +this long between reads. This reduces the load on the +.Ar source +device and the system in general. .It Fl p Ar pause .Xr sleep 3 this long whenever a read fails. This makes the diff --git a/sbin/recoverdisk/recoverdisk.c b/sbin/recoverdisk/recoverdisk.c index f13a1f211863..5971f78738ac 100644 --- a/sbin/recoverdisk/recoverdisk.c +++ b/sbin/recoverdisk/recoverdisk.c @@ -28,6 +28,11 @@ #include <time.h> #include <unistd.h> +/* + * This is a compromise between speed and wasted effort + */ +#define COMPROMISE_SIZE (128<<10) + struct lump { uint64_t start; uint64_t len; @@ -51,6 +56,7 @@ static uint64_t medium_read; static uint64_t small_read; static uint64_t total_size; static uint64_t done_size; +static uint64_t wasted_size; static char *input; static char *write_worklist_file = NULL; static char *read_worklist_file = NULL; @@ -61,6 +67,7 @@ static FILE *log_file = NULL; static char *work_buf; static char *pattern_buf; static double error_pause; +static double interval; static unsigned nlumps; static double n_reads, n_good_reads; @@ -418,7 +425,8 @@ fill_buf(char *buf, int64_t len, const char *pattern) static void usage(void) { - fprintf(stderr, "usage: recoverdisk [-b big_read] [-r readlist] " + fprintf(stderr, "usage: recoverdisk " + "[-b big_read] [-i interval ] [-r readlist] " "[-s interval] [-w writelist] source [destination]\n"); /* XXX update */ exit(1); @@ -486,6 +494,7 @@ attempt_one_lump(time_t t_now) fflush(log_file); } } else { + wasted_size += sz; printf("%14ju %7ju read error %d: (%s)", (uintmax_t)lp->start, (uintmax_t)sz, error, strerror(error)); @@ -557,8 +566,6 @@ determine_read_sizes(void) u_int sectorsize; off_t stripesize; - determine_total_size(); - #ifdef DIOCGSECTORSIZE if (small_read == 0) { error = ioctl(read_fd, DIOCGSECTORSIZE, §orsize); @@ -572,8 +579,8 @@ determine_read_sizes(void) #endif if (small_read == 0) { - printf("Assuming 512 for small_read\n"); small_read = 512; + printf("# Defaulting small_read to %ju\n", (uintmax_t)small_read); } if (medium_read && (medium_read % small_read)) { @@ -593,13 +600,13 @@ determine_read_sizes(void) #ifdef DIOCGSTRIPESIZE if (medium_read == 0) { error = ioctl(read_fd, DIOCGSTRIPESIZE, &stripesize); - if (error < 0 || stripesize < 0) { + if (error < 0 || stripesize <= 0) { // nope } else if ((uint64_t)stripesize < small_read) { // nope } else if (stripesize % small_read) { // nope - } else if (0 < stripesize && stripesize < (128<<10)) { + } else if (stripesize <= COMPROMISE_SIZE) { medium_read = stripesize; printf("# Got medium_read from DIOCGSTRIPESIZE: %ju\n", (uintmax_t)medium_read @@ -607,6 +614,7 @@ determine_read_sizes(void) } } #endif + #if defined(DIOCGFWSECTORS) && defined(DIOCGFWHEADS) if (medium_read == 0) { u_int fwsectors = 0, fwheads = 0; @@ -616,10 +624,16 @@ determine_read_sizes(void) error = ioctl(read_fd, DIOCGFWHEADS, &fwheads); if (error) fwheads = 0; - if (fwsectors && fwheads) { + if (fwsectors * fwheads * small_read <= COMPROMISE_SIZE) { medium_read = fwsectors * fwheads * small_read; printf( - "# Got medium_read from DIOCGFW{SECTORS,HEADS}: %ju\n", + "# Got medium_read from DIOCGFW{SECTORS*HEADS}: %ju\n", + (uintmax_t)medium_read + ); + } else if (fwsectors * small_read <= COMPROMISE_SIZE) { + medium_read = fwsectors * small_read; + printf( + "# Got medium_read from DIOCGFWSECTORS: %ju\n", (uintmax_t)medium_read ); } @@ -627,10 +641,11 @@ determine_read_sizes(void) #endif if (big_read == 0 && medium_read != 0) { - if (medium_read > (64<<10)) { + if (medium_read * 2 > COMPROMISE_SIZE) { big_read = medium_read; + medium_read = 0; } else { - big_read = 128 << 10; + big_read = COMPROMISE_SIZE; big_read -= big_read % medium_read; } printf("# Got big_read from medium_read: %ju\n", @@ -639,12 +654,16 @@ determine_read_sizes(void) } if (big_read == 0) { - big_read = 128 << 10; + big_read = COMPROMISE_SIZE; + big_read -= big_read % small_read; printf("# Defaulting big_read to %ju\n", (uintmax_t)big_read ); } + if (medium_read >= big_read) + medium_read = 0; + if (medium_read == 0) { /* * We do not want to go directly to single sectors, but @@ -662,12 +681,20 @@ determine_read_sizes(void) (uintmax_t)medium_read ); } - fprintf(stderr, - "# Bigsize = %ju, medium_read = %ju, small_read = %ju\n", + printf("# Bigsize = %ju, medium_read = %ju, small_read = %ju\n", (uintmax_t)big_read, (uintmax_t)medium_read, (uintmax_t)small_read); -} + assert(0 < small_read); + + assert(0 < medium_read); + assert(medium_read >= small_read); + assert(medium_read <= big_read); + assert(medium_read % small_read == 0); + assert(0 < big_read); + assert(big_read >= medium_read); + assert(big_read % small_read == 0); +} /**********************************************************************/ @@ -687,15 +714,14 @@ monitor_read_sizes(uint64_t failed_size) ); big_read = medium_read; medium_read = small_read; + wasted_size = 0; return; } - if (failed_size > small_read) { - if (n_reads < n_good_reads + 100) - return; + if (big_read > small_read && wasted_size / small_read > 200) { fprintf( stderr, - "Too many failures." + "Too much wasted effort." " (%.0f bad of %.0f)" " Shifting to small_reads.\n", n_reads - n_good_reads, n_reads @@ -719,11 +745,14 @@ main(int argc, char * const argv[]) setbuf(stdout, NULL); setbuf(stderr, NULL); - while ((ch = getopt(argc, argv, "b:l:p:m:r:w:s:t:u:v")) != -1) { + while ((ch = getopt(argc, argv, "b:i:l:p:m:r:w:s:t:u:v")) != -1) { switch (ch) { case 'b': big_read = strtoul(optarg, NULL, 0); break; + case 'i': + interval = strtod(optarg, NULL); + break; case 'l': log_file = fopen(optarg, "a"); if (log_file == NULL) { @@ -774,6 +803,8 @@ main(int argc, char * const argv[]) if (read_fd < 0) err(1, "Cannot open read descriptor %s", argv[0]); + determine_total_size(); + determine_read_sizes(); work_buf = malloc(big_read); @@ -816,6 +847,9 @@ main(int argc, char * const argv[]) t_save = t_first; unsaved = 0; while (!aborting) { + if (interval > 0) { + usleep((unsigned long)(1e6 * interval)); + } t_now = time(NULL); sz = attempt_one_lump(t_now); error = errno;
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?202509020549.5825nUbh082126>