htop memory usage error

[instain]

I've had htop installed on two different machines now, one running FreeBSD 6.3 and one 7.1. On both of them htop reports memory usage that is off by a factor of 1024.

Here's an example comparing top to htop

top:
PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND
61920 apaulsen 7 52 0 258M 199M ucond 0 0:00 0.59% firefox-bin

htop:
PID USER PRI NI VIRT RES SHR S CPU% MEM% TIME+ Command
61920 apaulsen 128 0 270608M 209856M 0 S 11.0 100. 23h24:07 /usr/local/lib/firefox/firefox-bin

Any ideas on how to fix this?

[instain]

I found the problem and attempted to push a patch upstream. In case anyone is interested:

--- Process.c~ 2008-09-22 22:43:34.000000000 -0500
+++ Process.c 2009-02-24 18:14:51.000000000 -0600
@@ -33,8 +33,9 @@
// This works only with glibc 2.1+. On earlier versions
// the behavior is similar to have a hardcoded page size.
#ifndef PAGE_SIZE
-#define PAGE_SIZE ( sysconf(_SC_PAGESIZE) / 1024 )
+#define PAGE_SIZE ( sysconf(_SC_PAGESIZE) )
#endif
+#define PAGE_SIZE_KB ( PAGE_SIZE / ONE_K )

#define PROCESS_COMM_LEN 300

@@ -351,13 +352,13 @@
: attr;
break;
}
- case M_DRS: Process_printLargeNumber(this, str, this->m_drs * PAGE_SIZE); return;
- case M_DT: Process_printLargeNumber(this, str, this->m_dt * PAGE_SIZE); return;
- case M_LRS: Process_printLargeNumber(this, str, this->m_lrs * PAGE_SIZE); return;
- case M_TRS: Process_printLargeNumber(this, str, this->m_trs * PAGE_SIZE); return;
- case M_SIZE: Process_printLargeNumber(this, str, this->m_size * PAGE_SIZE); return;
- case M_RESIDENT: Process_printLargeNumber(this, str, this->m_resident * PAGE_SIZE); return;
- case M_SHARE: Process_printLargeNumber(this, str, this->m_share * PAGE_SIZE); return;
+ case M_DRS: Process_printLargeNumber(this, str, this->m_drs * PAGE_SIZE_KB); return;
+ case M_DT: Process_printLargeNumber(this, str, this->m_dt * PAGE_SIZE_KB); return;
+ case M_LRS: Process_printLargeNumber(this, str, this->m_lrs * PAGE_SIZE_KB); return;
+ case M_TRS: Process_printLargeNumber(this, str, this->m_trs * PAGE_SIZE_KB); return;
+ case M_SIZE: Process_printLargeNumber(this, str, this->m_size * PAGE_SIZE_KB); return;
+ case M_RESIDENT: Process_printLargeNumber(this, str, this->m_resident * PAGE_SIZE_KB); return;
+ case M_SHARE: Process_printLargeNumber(this, str, this->m_share * PAGE_SIZE_KB); return;
case ST_UID: snprintf(buffer, n, "%4d ", this->st_uid); break;
case USER: {
if (Process_getuid != this->st_uid)

--- ProcessList.c~ 2008-09-23 01:23:14.000000000 -0500
+++ ProcessList.c 2009-02-24 18:14:13.000000000 -0600
@@ -698,7 +698,7 @@
period * 100.0;
process->percent_cpu = MAX(MIN(percent_cpu, processors*100.0), 0.0);

- process->percent_mem = (process->m_resident * PAGE_SIZE) /
+ process->percent_mem = (process->m_resident * PAGE_SIZE_KB) /
(float)(this->totalMem) *
100.0;

[vj]

Nice job!
Just discovered this wonderful utility, installed it and spotted the problem. Google leaded me to your post that have completely solved the problem.
Thanks!

[instain]

The patch was applied upstream so it should work properly on the next release.

[instain]

The patch was applied upstream so it should work properly on the next release.

Htop now reports memory correctly in version 0.8.2 (in ports already).