Skip to contentSkip to conversation

use display_size function instead of hardcoding

PATCH New Daniel Mierswa 3 participants 12 messages Sep 26, 2009 – Sep 30, 2009

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 use display_size function instead of hardcoding

We should use display_size everywhere for consistency reasons. Other commands and places that print the stripe/PE size already do that. This might break scripts that rely on the stripe/PE size being printed in KByte in those circumstances.
This requires changes in the .po file aswell.

---
 lib/display/display.c |    4 ++--
 lib/striped/striped.c |    4 ++--
 po/lvm2.po            |    4 ++--
 3 files changed, 6 insertions(+), 6 deletions(-)
lib/display/display.c +2 -2
index 62f30f6..b8da95b 100644
@@ -344,7 +344,7 @@ void pvdisplay_full(const struct cmd_context *cmd,
 	/* LV count is no longer available when displaying PV
 	   log_print("Cur LV                %u", vg->lv_count);
 	 */
-	log_print("PE Size (KByte)       %" PRIu32, pv->pe_size / 2);
+	log_print("PE Size               %s", display_size(cmd, (uint64_t) pv->pe_size));
 	log_print("Total PE              %u", pv->pe_count);
 	log_print("Free PE               %" PRIu32, pe_free);
 	log_print("Allocated PE          %u", pv->pe_alloc_count);
@@ -489,7 +489,7 @@ int lvdisplay_full(struct cmd_context *cmd,
 	log_print("Segments               %u", dm_list_size(&lv->segments));
 
 /********* FIXME Stripes & stripesize for each segment
-	log_print("Stripe size (KByte)    %u", lv->stripesize / 2);
+	log_print("Stripe size            %s", display_size(cmd, (uint64_t) lv->stripesize));
 ***********/
 
 	log_print("Allocation             %s", get_alloc_string(lv->alloc));
lib/striped/striped.c +2 -2
index 0e1ca2d..072e34b 100644
@@ -40,8 +40,8 @@ static void _striped_display(const struct lv_segment *seg)
 		display_stripe(seg, 0, "  ");
 	else {
 		log_print("  Stripes\t\t%u", seg->area_count);
-		log_print("  Stripe size\t\t%u KB", seg->stripe_size / 2);
-
+		log_print("  Stripe size\t\t%s", display_size(seg->lv->vg->cmd,
+		                                              (uint64_t) seg->stripe_size));
 		for (s = 0; s < seg->area_count; s++) {
 			log_print("  Stripe %d:", s);
 			display_stripe(seg, s, "    ");
po/lvm2.po +2 -2
index 2b236ae..8a14d7b 100644
@@ -1487,7 +1487,7 @@ msgstr ""
 
 #: display/display.c:312
 #, c-format
-msgid "PE Size (KByte)       %u"
+msgid "PE Size               %s"
 msgstr ""
 
 #: display/display.c:313 display/display.c:592
@@ -6510,7 +6510,7 @@ msgstr ""
 
 #: striped/striped.c:42
 #, c-format
-msgid "  Stripe size\t\t%u KB"
+msgid "  Stripe size\t\t%s"
 msgstr ""
 
 #: striped/striped.c:45
Signature · 1 line
1.6.4.4
#1 lore raw

PATCH 2/3 differentiate between S.I. and binary units in output

Current behaviour would print identical strings for S.I. and binary based units, say KB or Megabyte, also K/M/G in short notation. This can be confusing at times, because the manual for lvs/pvs/vgs etc. all state that you pass lowercase suffixes for binary and uppercase suffixes for S.I. suffixes.
This patch attempts to differentiate between those units in the output and print a proper string accordingly.
For binary based uints:
Long output strings (unsed) will be Kibibyte, Mebibyte, etc. Short output strings will be KiB, MiB, etc.
Unit output strings will be lower-cased: k, m, g, etc. Sectors determined by division with 512 will be suffixed with s.
For S.I. based units:
Long output strings (unused) will be Kilobyte, Megabyte, etc. Short output strings will be kB, MB, GB, etc.
Unit output strings will be upper-cased: K, M, G, etc. Sectors determined by division with 500 will be suffixed with S.

This will break every script that relies on the output of commands, which would print "KB", "MB", etc.
They need to be adjusted to respect above documented notions and also take into consideration that padding changed, i.e. " MB" = " MB ", since " MiB" and so forth have 3 letters. See the patch for details.

---
 lib/display/display.c |   54 ++++++++++++++++++++++++++++++++----------------
 1 files changed, 36 insertions(+), 18 deletions(-)
lib/display/display.c +36 -18
index b8da95b..1be7366 100644
@@ -156,20 +156,30 @@ static const char *_display_size(const struct cmd_context *cmd,
 	int s;
 	int suffix = 1, precision;
 	uint64_t byte = UINT64_C(0);
-	uint64_t units = UINT64_C(1024);
 	char *size_buf = NULL;
 	const char * const size_str[][3] = {
-		{" Exabyte", " EB", "E"},
-		{" Petabyte", " PB", "P"},
-		{" Terabyte", " TB", "T"},
-		{" Gigabyte", " GB", "G"},
-		{" Megabyte", " MB", "M"},
-		{" Kilobyte", " KB", "K"},
+		{" Exbibyte", " EiB", "e"},
+		{" Pebibyte", " PiB", "p"},
+		{" Tebibyte", " TiB", "t"},
+		{" Gibibyte", " GiB", "g"},
+		{" Mebibyte", " MiB", "m"},
+		{" Kibibyte", " KiB", "k"},
+		{" Sectors ", " Se ", "s"},
+
+		{" Exabyte ", " EB ", "E"},
+		{" Petabyte", " PB ", "P"},
+		{" Terabyte", " TB ", "T"},
+		{" Gigabyte", " GB ", "G"},
+		{" Megabyte", " MB ", "M"},
+		{" Kilobyte", " kB ", "K"},
+		{" Sectors ", " Se ", "S"},
+
 		{"", "", ""},
-		{" Byte    ", " B ", "B"},
-		{" Units   ", " Un", "U"},
-		{" Sectors ", " Se", "S"},
-		{"         ", "   ", " "},
+
+		{" Byte    ", " B  ", "B"},
+		{" Units   ", " Un ", "U"},
+
+		{"         ", "    ", " "},
 	};
 
 	if (!(size_buf = dm_pool_alloc(cmd->mem, SIZE_BUF))) {
@@ -179,9 +189,14 @@ static const char *_display_size(const struct cmd_context *cmd,
 
 	suffix = cmd->current_settings.suffix;
 
-	for (s = 0; s < 10; s++)
-		if (toupper((int) cmd->current_settings.unit_type) ==
-		    *size_str[s][2])
+	/* find one of kmgtpesKMGTPES */
+	for (s = 0; s < 14; s++)
+		if (cmd->current_settings.unit_type == *size_str[s][2])
+			break;
+
+	/* find one of bBuU */
+	for (; s >= 14 && s < 17; s++)
+		if (toupper((int) cmd->current_settings.unit_type) == *size_str[s][2])
 			break;
 
 	if (size == UINT64_C(0)) {
@@ -191,16 +206,19 @@ static const char *_display_size(const struct cmd_context *cmd,
 
 	size *= UINT64_C(512);
 
-	if (s < 10)
+	if (s < 17)
 		byte = cmd->current_settings.unit_factor;
 	else {
+		uint64_t units;
 		suffix = 1;
-		if (cmd->current_settings.unit_type == 'H')
+		if (cmd->current_settings.unit_type == 'H') {
+			s = 6;
 			units = UINT64_C(1000);
-		else
+		} else {
+			s = 0;
 			units = UINT64_C(1024);
+		}
 		byte = units * units * units * units * units * units;
-		s = 0;
 		while (size_str[s] && size < byte)
 			s++, byte /= units;
 	}
Signature · 1 line
1.6.4.4
#2 lore raw

PATCH 2/3 differentiate between S.I. and binary units in output

Quoted 10 lines · Daniel Mierswa

For binary based uints:
Long output strings (unsed) will be Kibibyte, Mebibyte, etc. Short output strings will be KiB, MiB, etc.
Unit output strings will be lower-cased: k, m, g, etc. Sectors determined by division with 512 will be suffixed with s.
For S.I. based units:
Long output strings (unused) will be Kilobyte, Megabyte, etc. Short output strings will be kB, MB, GB, etc.
Unit output strings will be upper-cased: K, M, G, etc. Sectors determined by division with 500 will be suffixed with S.

Bytes, Units and Sectors are special cases - in a separate group. A sector is always 512 bytes. I don't see any need to support 500-byte sectors.

As with the last patch, I can't put this into RHEL5 so it needs to be conditional (selectable at both compilation time and run time).

Alasdair.

#3 lore raw

Re: [PATCH 2/3] differentiate between S.I. and binary units in output

Quoted 12 lines · Alasdair G Kergon
Quoted 10 lines · Daniel Mierswa

For binary based uints:
Long output strings (unsed) will be Kibibyte, Mebibyte, etc. Short output strings will be KiB, MiB, etc.
Unit output strings will be lower-cased: k, m, g, etc. Sectors determined by division with 512 will be suffixed with s.
For S.I. based units:
Long output strings (unused) will be Kilobyte, Megabyte, etc. Short output strings will be kB, MB, GB, etc.
Unit output strings will be upper-cased: K, M, G, etc. Sectors determined by division with 500 will be suffixed with S.

Quoted 3 lines

Bytes, Units and Sectors are special cases - in a separate group. A sector is always 512 bytes. I don't see any need to support 500-byte sectors.

I agree that 500-byte sectors aren't meaningful but I added support for them a few months ago to be consistent, see:
http://sources.redhat.com/git/?p=lvm2.git;a=commit;h=69da2ac0

#4 lore raw

Re: [PATCH 2/3] differentiate between S.I. and binary units in output

Quoted 3 lines · Mike Snitzer

I agree that 500-byte sectors aren't meaningful but I added support for them a few months ago to be consistent, see:
http://sources.redhat.com/git/?p=lvm2.git;a=commit;h=69da2ac0

Where's that in WHATS_NEW?

I'll probably take it back out: I can't think of any reason for supporting it and confusing people.

Alasdair

#5 lore raw

Re: [PATCH 2/3] differentiate between S.I. and binary units in output

Quoted 9 lines · Alasdair G Kergon
Quoted 3 lines · Mike Snitzer

I agree that 500-byte sectors aren't meaningful but I added support for them a few months ago to be consistent, see:
http://sources.redhat.com/git/?p=lvm2.git;a=commit;h=69da2ac0

Where's that in WHATS_NEW?

I'll probably take it back out: I can't think of any reason for supporting it and confusing people.

Right, I agree that confusing people isn't good.. but I thought it better to allow people to specify 'S' without erroring out.

As for WHATS_NEW; I added the following to the 2.02.49 release:
Update the man pages to document size units uniformly. Allow commandline sizes to be specified in terms of bytes and sectors.

If you do back it out please still allow 's'; it is needed for pvcreate --dataalignmentoffset

But as we discussed, its probably best if both 's' and 'S' mean 512 bytes.

Mike

#6 lore raw

PATCH 2/3 differentiate between S.I. and binary units in output

In reply to Daniel Mierswa

So please would you:
1) Make this change conditional on a runtime check of
cmd->si_unit_consistency;
2) Remove the 500-byte sector offering.

The new-style output can be the default from 2.02.54 onwards - Fedora 13, RHEL6 etc.

People who find problems with their scripts after an upgrade can easily reinstate the old inconsistent output by using an lvm.conf setting.

Distros that can't change the output (e.g. RHEL5) can set a configure parameter.

Alasdair

#7 lore raw

PATCH 2/3 differentiate between S.I. and binary units in output

Quoted 3 lines · Alasdair G Kergon

So please would you:
1) Make this change conditional on a runtime check of
cmd->si_unit_consistency;

Attached, hope it's bug free, somehow all things went crazy and I got very busy this week but don't couldn't wait any longer to get this merged.

Quoted 1 line

2) Remove the 500-byte sector offering.

Figured, you already did.

Quoted 2 lines

The new-style output can be the default from 2.02.54 onwards - Fedora 13, RHEL6 etc.

We still need to tweak the documentation at places, which I can do as soon as the patches are for the actual code handling are applied.

Quoted 2 lines

People who find problems with their scripts after an upgrade can easily reinstate the old inconsistent output by using an lvm.conf setting.

Yep, win win. Good solution in the end anyway although it requires the LVM2 development team to keep "old" code around, which I prefer not to if I can avoid it. ;)

Signature · 9 lines
Mierswa, Daniel

If you still don't like it, that's ok: that's why I'm boss. I simply 
know better than you do.
                --- Linus Torvalds, comp.os.linux.advocacy, 1996/07/22
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 0001-differentiate-between-S.I.-and-binary-units-in-outpu.patch
URL: <http://listman.redhat.com/archives/lvm-devel/attachments/20090930/394a27a0/attachment.ksh>
#8 lore raw

PATCH 2/3 differentiate between S.I. and binary units in output

OK - I've applied this.

I added a 'base' value within the array and some #defines to try to make the code easier to understand. Hopefully I didn't introduce any bugs during that process.

Alasdair

#9 lore raw

PATCH 3/3 adjust the testsuite to work with new output

In reply to Daniel Mierswa

The last change made LVM tools print different size strings and the testsuite needs to be adjusted to recognize those.

---
 test/t-lvcreate-usage.sh |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
test/t-lvcreate-usage.sh +2 -2
index 741351d..4180a96 100755
@@ -30,7 +30,7 @@ grep "^  Negative stripesize is invalid\$" err
 
 # 'lvcreate rejects a too-large stripesize'
 not lvcreate -L 64M -n $lv -i2 --stripesize 4294967291 $vg 2>err
-grep "^  Stripe size cannot be larger than 512.00 GB\$" err
+grep "^  Stripe size cannot be larger than 512.00 GiB\$" err
 
 # 'lvcreate w/single stripe succeeds with diagnostics to stdout' 
 lvcreate -L 64M -n $lv -i1 --stripesize 4 $vg >out 2>err 
@@ -53,7 +53,7 @@ grep "^  Number of stripes (129) must be between 1 and 128\$" err
 # The case on lvdisplay output is to verify that the LV was not created.
 # 'lvcreate rejects an invalid stripe size'
 not lvcreate -L 64M -n $lv -i2 --stripesize 3 $vg 2>err
-grep "^  Invalid stripe size 3\.00 KB\$" err
+grep "^  Invalid stripe size 3\.00 KiB\$" err
 case $(lvdisplay $vg) in "") true ;; *) false ;; esac
 
 # Setting max_lv works. (bz490298)
Signature · 1 line
1.6.4.4
#10 lore raw

PATCH 1/3 use display_size function instead of hardcoding

In reply to Daniel Mierswa
Quoted 4 lines · Daniel Mierswa

We should use display_size everywhere for consistency reasons. Other commands and places that print the stripe/PE size already do that. This might break scripts that rely on the stripe/PE size being printed in KByte in those circumstances.

I can't make this change in RHEL5 and since we use the upstream code base that means we need to make it conditional, probably with an lvm.conf setting with a default value chosen at configuration time.

Alasdair