[Bug 5657] EVE Online reports VM Size as 0 while on Windows it reports the memory usage (NtQueryInformationProcess ProcessVmCounters info class: actual value for 'PagefileUsage' required)
wine-bugs at winehq.org
wine-bugs at winehq.org
Sun Oct 27 16:02:50 CDT 2013
http://bugs.winehq.org/show_bug.cgi?id=5657
Anastasius Focht <focht at gmx.net> changed:
What |Removed |Added
----------------------------------------------------------------------------
URL| |http://community.eveonline.
| |com/support/download/
CC| |focht at gmx.net
Component|-unknown |ntdll
Version|unspecified |0.9.17.
Summary|EVE Online reports VM Size |EVE Online reports VM Size
|as 0 while on Windows it |as 0 while on Windows it
|reports the memory usage |reports the memory usage
| |(NtQueryInformationProcess
| |ProcessVmCounters info
| |class: actual value for
| |'PagefileUsage' required)
--- Comment #13 from Anastasius Focht <focht at gmx.net> 2013-10-27 16:02:50 CDT ---
Hello folks,
is seems the culprit is NtQueryInformationProcess -> ProcessVmCounters not
returning useful information.
There are some tools to reverse EVE Online client code (compiled stackless
python and encrypted).
You need:
http://python.org/ftp/python/2.7.5/python-2.7.5.msi (Python for win32)
https://github.com/wibiti/evedec (decrypts Eve Online python files and passes
them to uncompyle2 to decompile)
https://github.com/wibiti/uncompyle2 (Python 2.7 byte-code decompiler, written
in Python 2.7)
Unfortunately 'evedec' makes use of parallel processing lib, requiring pipe
message mode (bug 17195 and bug 17273) so you have to work around with
unofficial patchset(s) floating around or rewriting the tool to use single
processing mode (without Queue).
Decompiling eve-8.32.622857 client and looking for code that updates the memory
stat page gives a hit here:
carbon/client/script/util/monitor.py:
--- snip ---
...
def UpdateUI(self):
wnd = self.GetWnd()
if not wnd or wnd.destroyed:
return
hd, li = blue.pyos.ProbeStuff()
info = Row(list(hd), li)
virtualMem = info.pagefileUsage / 1024
if self.lastVM is None:
self.lastVM = virtualMem
delta = virtualMem - self.lastVM
self.totalVMDelta += delta
self.lastVM = virtualMem
delta = delta or self.lastVMDelta
self.lastVMDelta = delta
dc = ['<color=0xff00ff00>', '<color=0xffff0000>'][delta > 0]
tdc = ['<color=0xff00ff00>', '<color=0xffff0000>'][self.totalVMDelta >
0]
try:
dev = trinity.device
iml = ''
if blue.logInMemory.isActive:
iml = 'In-memory logging <color=0xff00ff00>active</color> (%s /
%s)' % (LOGTYPE_MAPPING.get(blue.logInMemory.threshold, ('Unknown', ''))[0],
blue.logInMemory.capacity)
fps = 'Fps: %.1f - %s' % (blue.os.fps, iml)
if wnd.sr.fpsText.text != fps:
wnd.sr.fpsText.text = fps
vm = 'VM Size/D/TD: %sK / %s%sK<color=0xffffffff> /
%s%sK<color=0xffb0b0b0> - totalVidMem: %sM' % (util.FmtAmt(virtualMem),
dc,
util.FmtAmt(delta),
tdc,
util.FmtAmt(self.totalVMDelta),
self.totalVidMem)
if wnd.sr.vmText.text != vm:
wnd.sr.vmText.text = vm
inUse = util.FmtAmt(blue.motherLode.memUsage / 1024)
total = util.FmtAmt(blue.motherLode.maxMemUsage / 1024)
num = blue.motherLode.size()
vm = 'Resource Cache Usage: %sK / %sK - %s objects' % (inUse,
total, num)
if wnd.sr.cacheText.text != vm:
wnd.sr.cacheText.text = vm
spaceMgr = sm.StartService('space')
mo = 'Lazy Queue: %s' % getattr(spaceMgr, 'lazyLoadQueueCount', 0)
if session.role & service.ROLEMASK_ELEVATEDPLAYER:
mo += ' - Preload Queue: %s' % getattr(spaceMgr,
'preloadQueueCount', 22)
if wnd.sr.queueText.text != mo:
wnd.sr.queueText.text = mo
except Exception as e:
print 'e', e
self.uitimer = None
sys.exc_clear()
...
--- snip ---
The value in question is 'info.pagefileUsage' (the others are just interval
deltas).
carbon/common/lib/win32.py:
--- snip ---
...
class PROCESS_MEMORY_COUNTERS_EX(Structure):
_fields_ = [('cb', DWORD),
('PageFaultCount', DWORD),
('PeakWorkingSetSize', SIZE_T),
('WorkingSetSize', SIZE_T),
('QuotaPeakPagedPoolUsage', SIZE_T),
('QuotaPagedPoolUsage', SIZE_T),
('QuotaPeakNonPagedPoolUsage', SIZE_T),
('QuotaNonPagedPoolUsage', SIZE_T),
('PagefileUsage', SIZE_T),
('PeakPagefileUsage', SIZE_T),
('PrivateUsage', SIZE_T)]
...
def GetProcessMemoryInfo():
f = windll.Kernel32.GetCurrentProcess
f.retval = HANDLE
h = f()
counters = PROCESS_MEMORY_COUNTERS_EX()
windll.Psapi.GetProcessMemoryInfo(HANDLE(h), byref(counters),
sizeof(counters))
return StructToKeyval(counters)
...
--- snip ---
Other code making use of that value:
carbon/common/script/net/ExceptionWrapperGPCS.py:
--- snip ---
ram = blue.win32.GetProcessMemoryInfo()['PagefileUsage'] / 1024 / 1024
cpuLoad = self.machoNet.GetCPULoad()
m = blue.win32.GlobalMemoryStatus()
memLeft = m['AvailPhys'] / 1024 / 1024
txt = 'System Information: '
txt += 'Total CPU load: %s%%' % int(cpuLoad)
txt += ' | Process memory in use: %s MB' % ram
txt += ' | Physical memory left: %s MB\n' % memLeft
--- snip ---
Psapi.GetProcessMemoryInfo -> kernel32.K32GetProcessMemoryInfo ->
NtQueryInformationProcess(process, ProcessVmCounters, ...)
I found the following articles useful how to estimate/substitute some values:
http://stackoverflow.com/questions/372484/how-do-i-programmatically-check-memory-use-in-a-fairly-portable-way-c-c
http://nadeausoftware.com/articles/2012/07/c_c_tip_how_get_process_resident_set_size_physical_memory_use
http://locklessinc.com/articles/memory_usage/
As already mentioned, only 'PagefileUsage' is actually required for EVE.
The best substitute for 'PagefileUsage' might be 'VmSize' from either
'/proc/<pid>/stat' (VmSize) or '/proc/<pid>/statm' (VmSize*page_size).
I added a small Linux implementation to ProcessVmCounters getter to fill
'pvmi.PagefileUsage' member with current VmSize and it was enough to have EVE
client display proper memory usage (current + deltas).
$ wine --version
wine-1.7.5-100-g3ab406d
Regards
--
Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email
Do not reply to this email, post in Bugzilla using the
above URL to reply.
------- You are receiving this mail because: -------
You are watching all bug changes.
More information about the wine-bugs
mailing list