From: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>
To: Christian Borntraeger <borntraeger@de.ibm.com>,
Paolo Bonzini <pbonzini@redhat.com>,
"qemu-devel@nongnu.org" <qemu-devel@nongnu.org>,
"qemu-block@nongnu.org" <qemu-block@nongnu.org>
Cc: "kvm@vger.kernel.org" <kvm@vger.kernel.org>,
"berto@igalia.com" <berto@igalia.com>,
"mdroth@linux.vnet.ibm.com" <mdroth@linux.vnet.ibm.com>,
"armbru@redhat.com" <armbru@redhat.com>,
"ehabkost@redhat.com" <ehabkost@redhat.com>,
"rth@twiddle.net" <rth@twiddle.net>,
"mtosatti@redhat.com" <mtosatti@redhat.com>,
Denis Lunev <den@virtuozzo.com>,
Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Subject: Re: [PATCH 3/3] i386/kvm: initialize struct at full before ioctl call
Date: Wed, 31 Jul 2019 14:11:02 +0000 [thread overview]
Message-ID: <a38ae11d-4dee-d50b-7719-e65c5564a985@virtuozzo.com> (raw)
In-Reply-To: <f9346216-a4e9-4882-4a36-33580529b75e@de.ibm.com>
On 31/07/2019 15:43, Christian Borntraeger wrote:
>
>
> On 31.07.19 14:28, Christian Borntraeger wrote:
>>
>>
>> On 31.07.19 14:04, Andrey Shinkevich wrote:
>>> On 31/07/2019 10:24, Christian Borntraeger wrote:
>>>>
>>>>
>>>> On 30.07.19 21:20, Paolo Bonzini wrote:
>>>>> On 30/07/19 18:01, Andrey Shinkevich wrote:
>>>>>> Not the whole structure is initialized before passing it to the KVM.
>>>>>> Reduce the number of Valgrind reports.
>>>>>>
>>>>>> Signed-off-by: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>
>>>>>
>>>>> Christian, is this the right fix? It's not expensive so it wouldn't be
>>>>> an issue, just checking if there's any better alternative.
>>>>
>>>> I think all of these variants are valid with pros and cons
>>>> 1. teach valgrind about this:
>>>> Add to coregrind/m_syswrap/syswrap-linux.c (and the relevant header files)
>>>> knowledge about which parts are actually touched.
>>>> 2. use designated initializers
>>>> 3. use memset
>>>> 3. use a valgrind callback VG_USERREQ__MAKE_MEM_DEFINED to tell that this memory is defined
>>>>
>>>
>>> Thank you all very much for taking part in the discussion.
>>> Also, one may use the Valgrind technology to suppress the unwanted
>>> reports by adding the Valgrind specific format file valgrind.supp to the
>>> QEMU project. The file content is extendable for future needs.
>>> All the cases we like to suppress will be recounted in that file.
>>> A case looks like the stack fragments. For instance, from QEMU block:
>>>
>>> {
>>> hw/block/hd-geometry.c
>>> Memcheck:Cond
>>> fun:guess_disk_lchs
>>> fun:hd_geometry_guess
>>> fun:blkconf_geometry
>>> ...
>>> fun:device_set_realized
>>> fun:property_set_bool
>>> fun:object_property_set
>>> fun:object_property_set_qobject
>>> fun:object_property_set_bool
>>> }
>>>
>>> The number of suppressed cases are reported by the Valgrind with every
>>> run: "ERROR SUMMARY: 5 errors from 3 contexts (suppressed: 0 from 0)"
>>>
>>> Andrey
>>
>> Yes, indeed that would be another variant. How performance critical are
>> the fixed locations? That might have an impact on what is the best solution.
>> From a cleanliness approach doing 1 (adding the ioctl definition to valgrind)
>> is certainly the most beautiful way. I did that in the past, look for example at
>>
>> https://sourceware.org/git/?p=valgrind.git;a=commitdiff;h=c2baee9b7bf043702c130de0771a4df439fcf403
>> or
>> https://sourceware.org/git/?p=valgrind.git;a=commitdiff;h=00a31dd3d1e7101b331c2c83fca6c666ba35d910
>>
>> for examples.
>>
>>
>>>
>>>>>
>>>>> Paolo
>>>>>
>>>>>> ---
>>>>>> target/i386/kvm.c | 3 +++
>>>>>> 1 file changed, 3 insertions(+)
>>>>>>
>>>>>> diff --git a/target/i386/kvm.c b/target/i386/kvm.c
>>>>>> index dbbb137..ed57e31 100644
>>>>>> --- a/target/i386/kvm.c
>>>>>> +++ b/target/i386/kvm.c
>>>>>> @@ -190,6 +190,7 @@ static int kvm_get_tsc(CPUState *cs)
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> + memset(&msr_data, 0, sizeof(msr_data));
>>>>>> msr_data.info.nmsrs = 1;
>>>>>> msr_data.entries[0].index = MSR_IA32_TSC;
>>>>>> env->tsc_valid = !runstate_is_running();
>>>>>> @@ -1706,6 +1707,7 @@ int kvm_arch_init_vcpu(CPUState *cs)
>>>>>>
>>>>>> if (has_xsave) {
>>>>>> env->xsave_buf = qemu_memalign(4096, sizeof(struct kvm_xsave));
>>>>>> + memset(env->xsave_buf, 0, sizeof(struct kvm_xsave));
>
> This is memsetting 4k?
> Yet another variant would be to use the RUNNING_ON_VALGRIND macro from
> valgrind/valgrind.h to only memset for valgrind. But just using MAKE_MEM_DEFINED
> from memcheck.h is simpler.
>
So, on this assumption, the code would look like
#ifdef CONFIG_VALGRIND_H
#include <valgrind/memcheck.h>
#endif
#ifdef CONFIG_VALGRIND_H
VALGRIND_MAKE_MEM_DEFINED(&msr_data, sizeof(msr_data));
#endif
etc.
Andrey
--
With the best regards,
Andrey Shinkevich
next prev parent reply other threads:[~2019-07-31 14:11 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-07-30 16:01 [PATCH 0/3] Reduce the number of Valgrind reports in unit tests Andrey Shinkevich
2019-07-30 16:01 ` [PATCH 1/3] test-throttle: Fix uninitialized use of burst_length Andrey Shinkevich
2019-08-13 12:19 ` Alberto Garcia
2019-07-30 16:01 ` [PATCH 2/3] tests: Fix uninitialized byte in test_visitor_in_fuzz Andrey Shinkevich
2019-07-30 16:01 ` [PATCH 3/3] i386/kvm: initialize struct at full before ioctl call Andrey Shinkevich
2019-07-30 16:44 ` [Qemu-devel] " Philippe Mathieu-Daudé
2019-07-30 17:05 ` Christian Borntraeger
2019-07-30 17:14 ` Philippe Mathieu-Daudé
2019-07-30 17:47 ` Christian Borntraeger
2019-07-31 9:05 ` Christophe de Dinechin
2019-07-31 12:32 ` Paolo Bonzini
2019-07-31 14:10 ` Andrey Shinkevich
2019-07-30 19:22 ` Paolo Bonzini
2019-07-30 16:46 ` Peter Maydell
2019-07-30 17:09 ` Christian Borntraeger
2019-07-30 19:20 ` Paolo Bonzini
2019-07-31 7:24 ` Christian Borntraeger
2019-07-31 12:04 ` Andrey Shinkevich
2019-07-31 12:28 ` Christian Borntraeger
2019-07-31 12:43 ` Christian Borntraeger
2019-07-31 13:03 ` Paolo Bonzini
2019-07-31 14:11 ` Andrey Shinkevich [this message]
2019-08-13 12:02 ` [PATCH 0/3] Reduce the number of Valgrind reports in unit tests Andrey Shinkevich
2019-08-13 12:05 ` Paolo Bonzini
2019-08-13 12:08 ` Andrey Shinkevich
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=a38ae11d-4dee-d50b-7719-e65c5564a985@virtuozzo.com \
--to=andrey.shinkevich@virtuozzo.com \
--cc=armbru@redhat.com \
--cc=berto@igalia.com \
--cc=borntraeger@de.ibm.com \
--cc=den@virtuozzo.com \
--cc=ehabkost@redhat.com \
--cc=kvm@vger.kernel.org \
--cc=mdroth@linux.vnet.ibm.com \
--cc=mtosatti@redhat.com \
--cc=pbonzini@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=rth@twiddle.net \
--cc=vsementsov@virtuozzo.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).