From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:44490) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ZWfOf-00061u-Js for qemu-devel@nongnu.org; Tue, 01 Sep 2015 02:50:22 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1ZWfOb-00012k-IJ for qemu-devel@nongnu.org; Tue, 01 Sep 2015 02:50:21 -0400 Received: from e28smtp03.in.ibm.com ([122.248.162.3]:39145) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ZWfOa-00012V-Tm for qemu-devel@nongnu.org; Tue, 01 Sep 2015 02:50:17 -0400 Received: from /spool/local by e28smtp03.in.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Tue, 1 Sep 2015 12:20:11 +0530 Message-ID: <55E54A9F.5000203@linux.vnet.ibm.com> Date: Tue, 01 Sep 2015 12:20:07 +0530 From: Hemant Kumar MIME-Version: 1.0 References: <1440992554-5200-1-git-send-email-hemant@linux.vnet.ibm.com> <55E4B77C.1080801@redhat.com> In-Reply-To: <55E4B77C.1080801@redhat.com> Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH][RESEND] qemu/kvm_stat: Fix I/O error from kvm_stat List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Wei Huang , qemu-devel@nongnu.org Cc: qemu-ppc@nongnu.org, agraf@suse.de Hi Wei, Thanks for the review. On 09/01/2015 01:52 AM, Wei Huang wrote: > On 08/30/2015 10:42 PM, Hemant Kumar wrote: >> Running kvm_stat on a powerpc macine where the kernel is compiled with >> KVM_BOOK3S_HV_EXIT_TIMING config option, generates the following error : >> >> # kvm_stat >> >> Traceback (most recent call last): >> File "/usr/bin/kvm_stat", line 644, in >> curses.wrapper(tui, stats) >> File "/usr/lib64/python2.7/curses/wrapper.py", line 43, in wrapper >> return func(stdscr, *args, **kwds) >> File "/usr/bin/kvm_stat", line 550, in tui >> refresh(sleeptime) >> File "/usr/bin/kvm_stat", line 526, in refresh >> s = stats.get() >> File "/usr/bin/kvm_stat", line 486, in get >> new = d.read() >> File "/usr/bin/kvm_stat", line 29, in read >> return dict([(key, val(key)) for key in self._fields]) >> File "/usr/bin/kvm_stat", line 28, in val >> return int(file(self.base + '/' + key).read()) >> IOError: [Errno 21] Is a directory: '/sys/kernel/debug/kvm/vm12840' >> >> The error says that its trying to read the contents of a directory as if >> it were a file. The kernel here creates directories inside >> "/sys/kernel/debug/kvm/" for the individual VMs running on the >> machine. These directories contain timing information (for rm_entry, >> rm_intr, rm_exit, guest, cede) related to each vcpu of each VM. All of >> this is enabled with KVM_BOOK3S_HV_EXIT_TIMING config option for the >> kernel. > A suggestion: would it be useful to display these statistics in > perf_stat? I understand that these counters are per-vm-based; but they > might provide useful insights to end-users. It would be definitely useful to display these statistics, but since this script doesn't work on powerpc with the config option enabled, this patch should work as a fix for now. We can work on enhancing kvm_stat to display these per vm statistics. >> kvm_stat script isn't supposed to parse/read these directories and isn't >> even expecting any directories inside "/sys/kernel/debug/kvm/". This >> patch fixes the problem by avoiding lookup for these directories. >> >> Reported-by: Krishnaja Balachandran >> Signed-off-by: Hemant Kumar >> --- >> scripts/kvm/kvm_stat | 6 +++++- >> 1 file changed, 5 insertions(+), 1 deletion(-) >> >> diff --git a/scripts/kvm/kvm_stat b/scripts/kvm/kvm_stat >> index 7e5d256..1da7a53 100755 >> --- a/scripts/kvm/kvm_stat >> +++ b/scripts/kvm/kvm_stat >> @@ -18,7 +18,11 @@ from ctypes import * >> class DebugfsProvider(object): >> def __init__(self): >> self.base = '/sys/kernel/debug/kvm' >> - self._fields = os.listdir(self.base) >> + files = os.listdir(self.base) >> + self._fields = [] >> + for f in files: >> + if os.path.isfile(self.base + '/' + f): >> + self._fields.append(f) >> def fields(self): >> return self._fields >> def select(self, fields): >> > I tested this patch on x86 box and didn't see any problem. So: > > Reviewed-by: Wei Huang Thanks! -- Hemant Kumar