which walks the PTEs to check the PFNs.
> prot_none_pte_entry() gets the bogus PFN from pte_pfn() and returns
> -EACCES because it thinks mprotect() is trying to adjust a high MMIO
> address.
Looks good to me. You're right that case was missed.
Reviewed-by: Andi Kleen
I think Thomas is still on vacation, copying Linus.
-Andi
From: Andi Kleen
The stable backport of the
x86/speculation/l1tf: Disallow non privileged high MMIO PROT_NONE mappings
patch for 4.4 and 4.9 put new C code for !__HAVE_ARCH_PFN_MODIFY_ALLOWED
code outside the assembler ifdef. This breaks the xtensa and ia64
build as reported by 0day which
From: Andi Kleen
The stable backport of the
x86/speculation/l1tf: Disallow non privileged high MMIO PROT_NONE mappings
patch for 4.4 and 4.9 put new C code for !__HAVE_ARCH_PFN_MODIFY_ALLOWED
code outside the assembler ifdef. This breaks the xtensa and ia64
build as reported by 0day which
cause we can't access LDLAT msr
>
> however that's not so easy/straight forward to check,
> so I thought we could use check on PEBs instead, which
> needs to be there for mem events as well and seems to
> be enough to check
>
> thoughts? thanks,
> jirka
Acked-by: Andi Kleen
-Andi
cause we can't access LDLAT msr
>
> however that's not so easy/straight forward to check,
> so I thought we could use check on PEBs instead, which
> needs to be there for mem events as well and seems to
> be enough to check
>
> thoughts? thanks,
> jirka
Acked-by: Andi Kleen
-Andi
> Which simply does not work. Look at Goldmont Fam 6 Model 5C. The SoCs
> with that Fam/Model combination are:
>
> - Apollo Lake
> - Broxton (has two platforms: Morganfield and Willowtrail)
Right pick one. The others are the same for software purposes
and can be handled in the same way.
>
>
> Which simply does not work. Look at Goldmont Fam 6 Model 5C. The SoCs
> with that Fam/Model combination are:
>
> - Apollo Lake
> - Broxton (has two platforms: Morganfield and Willowtrail)
Right pick one. The others are the same for software purposes
and can be handled in the same way.
>
>
On Tue, Aug 07, 2018 at 07:48:51PM +0200, Peter Zijlstra wrote:
> On Tue, Aug 07, 2018 at 10:35:42AM -0700, Dave Hansen wrote:
> > On 08/07/2018 10:17 AM, kan.li...@linux.intel.com wrote:
> > > Denverton and Gemini Lake are platform names and should not be used for
> > > Processor Family stuff.
On Tue, Aug 07, 2018 at 07:48:51PM +0200, Peter Zijlstra wrote:
> On Tue, Aug 07, 2018 at 10:35:42AM -0700, Dave Hansen wrote:
> > On 08/07/2018 10:17 AM, kan.li...@linux.intel.com wrote:
> > > Denverton and Gemini Lake are platform names and should not be used for
> > > Processor Family stuff.
On Tue, Aug 07, 2018 at 10:17:27AM -0700, kan.li...@linux.intel.com wrote:
> From: Kan Liang
>
> Denverton and Gemini Lake are platform names and should not be used for
> Processor Family stuff. The microarchitecture codename should be used.
>
> DENVERTON is Goldmont server SoC. Rename
On Tue, Aug 07, 2018 at 10:17:27AM -0700, kan.li...@linux.intel.com wrote:
> From: Kan Liang
>
> Denverton and Gemini Lake are platform names and should not be used for
> Processor Family stuff. The microarchitecture codename should be used.
>
> DENVERTON is Goldmont server SoC. Rename
On Mon, Aug 06, 2018 at 08:35:15PM +0200, Peter Zijlstra wrote:
> > +static bool disable_counter_freezing;
> > +module_param(disable_counter_freezing, bool, 0444);
> > +MODULE_PARM_DESC(disable_counter_freezing, "Disable counter freezing
> > feature."
> > + "The PMI handler will fall
On Mon, Aug 06, 2018 at 08:35:15PM +0200, Peter Zijlstra wrote:
> > +static bool disable_counter_freezing;
> > +module_param(disable_counter_freezing, bool, 0444);
> > +MODULE_PARM_DESC(disable_counter_freezing, "Disable counter freezing
> > feature."
> > + "The PMI handler will fall
ile doing that also add -mrecord-mcount to CC_FLAGS_FTRACE (if gcc
> actually supports it).
>
> Signed-off-by: Vasily Gorbik
Acked-by: Andi Kleen
-Andi
ile doing that also add -mrecord-mcount to CC_FLAGS_FTRACE (if gcc
> actually supports it).
>
> Signed-off-by: Vasily Gorbik
Acked-by: Andi Kleen
-Andi
> [ Goes around and rummages ]
>
> Oh, never mind, looking around reminded me why: we want to map the
> kernel text in the top 31 bits, so that we can use the faster
> -mcmodel=kernel because all symbols fit in sign-extended 32 bits.
>
> Maybe there was some other reason too, but I think that's
> [ Goes around and rummages ]
>
> Oh, never mind, looking around reminded me why: we want to map the
> kernel text in the top 31 bits, so that we can use the faster
> -mcmodel=kernel because all symbols fit in sign-extended 32 bits.
>
> Maybe there was some other reason too, but I think that's
>The semantic difference being, as far as I can tell, that the "GT"
>counters do not sample randomly, but will rather offer a precise count of
>loads meeting the stated condition.Â
It's all the same underlying hardware event. The only difference is what
threshold
is used for
>The semantic difference being, as far as I can tell, that the "GT"
>counters do not sample randomly, but will rather offer a precise count of
>loads meeting the stated condition.Â
It's all the same underlying hardware event. The only difference is what
threshold
is used for
> now need to dive in deep into the actual implementation. Simple question: is
> this a bug?
These are derived events from MEM_TRANS_RETIRED.LOAD_LATENCY, which is in
the SDM. It's not a bug. In general the event lists used by perf are
newer versions than what is in the SDM.
-Andi
> now need to dive in deep into the actual implementation. Simple question: is
> this a bug?
These are derived events from MEM_TRANS_RETIRED.LOAD_LATENCY, which is in
the SDM. It's not a bug. In general the event lists used by perf are
newer versions than what is in the SDM.
-Andi
"Metzger, Markus T" writes:
Adding Dave Hansen
> Hello,
>
> Starting with 4.15 I noticed that BTS is sporadically missing the tail
> of the trace in the perf_event data buffer. It shows as
>
> [decode error (1): instruction overflow]
>
> in GDB. Chances to see this are higher the longer
"Metzger, Markus T" writes:
Adding Dave Hansen
> Hello,
>
> Starting with 4.15 I noticed that BTS is sporadically missing the tail
> of the trace in the perf_event data buffer. It shows as
>
> [decode error (1): instruction overflow]
>
> in GDB. Chances to see this are higher the longer
> At least we need to make user aware about risk of setting custom flags.
There are valid use cases to override the flags. I use it sometimes too,
and know some other people do to.
But you need to know what you're doing.
Perhaps a warning during build would be reasonable. So if you ask
for a
> At least we need to make user aware about risk of setting custom flags.
There are valid use cases to override the flags. I use it sometimes too,
and know some other people do to.
But you need to know what you're doing.
Perhaps a warning during build would be reasonable. So if you ask
for a
On Tue, Jul 03, 2018 at 11:26:09PM +0300, Kirill A. Shutemov wrote:
> On Tue, Jul 03, 2018 at 11:03:07AM -0700, Andi Kleen wrote:
> > On Tue, Jul 03, 2018 at 05:21:50PM +0300, Kirill A. Shutemov wrote:
> > > On Tue, Jul 03, 2018 at 03:44:03PM +0300, Kirill A. Shutemov wrote:
>
On Tue, Jul 03, 2018 at 11:26:09PM +0300, Kirill A. Shutemov wrote:
> On Tue, Jul 03, 2018 at 11:03:07AM -0700, Andi Kleen wrote:
> > On Tue, Jul 03, 2018 at 05:21:50PM +0300, Kirill A. Shutemov wrote:
> > > On Tue, Jul 03, 2018 at 03:44:03PM +0300, Kirill A. Shutemov wrote:
>
On Tue, Jul 03, 2018 at 05:21:50PM +0300, Kirill A. Shutemov wrote:
> On Tue, Jul 03, 2018 at 03:44:03PM +0300, Kirill A. Shutemov wrote:
> > On Tue, Jul 03, 2018 at 01:24:49PM +0200, Gabriel C wrote:
> > > 2018-07-01 23:32 GMT+02:00 Benjamin Gilbert :
> > > > On Sun, Jul 01, 2018 at 05:15:59PM
On Tue, Jul 03, 2018 at 05:21:50PM +0300, Kirill A. Shutemov wrote:
> On Tue, Jul 03, 2018 at 03:44:03PM +0300, Kirill A. Shutemov wrote:
> > On Tue, Jul 03, 2018 at 01:24:49PM +0200, Gabriel C wrote:
> > > 2018-07-01 23:32 GMT+02:00 Benjamin Gilbert :
> > > > On Sun, Jul 01, 2018 at 05:15:59PM
>
> So I think you're good... But yes, you raise an interresting point.
So it sounds like architectures that don't have an instruction atomic u64
*_user need to disable interrupts during the access, and somehow handle that
case when a page fault happens?
-Andi
>
> So I think you're good... But yes, you raise an interresting point.
So it sounds like architectures that don't have an instruction atomic u64
*_user need to disable interrupts during the access, and somehow handle that
case when a page fault happens?
-Andi
;
> > + if (match_metric(pe->metric_group, metric) ||
> > + match_metric(pe->metric_name, metric))
Why both the group and the metric?
I would just match the metric_name
Otherwise it's impossible to have a group called "transaction"
With that fixed it's ok for me.
Acked-by: Andi Kleen
-Andi
;
> > + if (match_metric(pe->metric_group, metric) ||
> > + match_metric(pe->metric_name, metric))
Why both the group and the metric?
I would just match the metric_name
Otherwise it's impossible to have a group called "transaction"
With that fixed it's ok for me.
Acked-by: Andi Kleen
-Andi
> >
> > You may need to add support for wildcard pmus to it.
> >
> > -Andi
> >
>
> Currently only one PMU on s390 does support transactions, its the PMU with
> counters
> named cpum_cf.
Right, but you don't want to hard code that pmu into builtin-stat.
-Andi
> >
> > You may need to add support for wildcard pmus to it.
> >
> > -Andi
> >
>
> Currently only one PMU on s390 does support transactions, its the PMU with
> counters
> named cpum_cf.
Right, but you don't want to hard code that pmu into builtin-stat.
-Andi
Thomas Richter writes:
>
> + /* Handle -T as -M transaction. Once platform specific metrics
> + * support has been added to the json files, all archiectures
> + * will use this approach.
> + */
> + if (!strcmp(perf_env__arch(NULL),
Thomas Richter writes:
>
> + /* Handle -T as -M transaction. Once platform specific metrics
> + * support has been added to the json files, all archiectures
> + * will use this approach.
> + */
> + if (!strcmp(perf_env__arch(NULL),
> In particular 1) means that any extra instructions executed/not executed
> will cause a replay divergence (in practice rr uses retired conditional
> branches rather than instructions, because the instruction counter is
> not accurate, while the branch one is). This alone causes a problem
> for
> In particular 1) means that any extra instructions executed/not executed
> will cause a replay divergence (in practice rr uses retired conditional
> branches rather than instructions, because the instruction counter is
> not accurate, while the branch one is). This alone causes a problem
> for
> Unfortunately, that is insufficient. Almost difference in CPU behavior
> between the replayer
> and the replayee. In particular, the problems for rr here are
> 1) xgetbv, which can introduce differing register contents (and if
> code branches on this,
> potentially differences in control
> Unfortunately, that is insufficient. Almost difference in CPU behavior
> between the replayer
> and the replayee. In particular, the problems for rr here are
> 1) xgetbv, which can introduce differing register contents (and if
> code branches on this,
> potentially differences in control
Patch seems pointless if you can already control CPUID, which rr
supports. Just disable AVX512 in CPUID. All code using AVX should check
cpuid (or will fail anyways).
-Andi
Patch seems pointless if you can already control CPUID, which rr
supports. Just disable AVX512 in CPUID. All code using AVX should check
cpuid (or will fail anyways).
-Andi
namic ftrace"),
> added a mismatched endif. This causes cmd_objtool to get mistakenly
> set.
>
> Relocate endif to balance the newly added -record-mcount check.
>
> Fixes: 96f60dfa5819 ("trace: Use -mcount-record for dynamic ftrace")
> Signed-off-by: Greg Thelen
Looks good thanks.
Acked-by: Andi Kleen
-Andi
namic ftrace"),
> added a mismatched endif. This causes cmd_objtool to get mistakenly
> set.
>
> Relocate endif to balance the newly added -record-mcount check.
>
> Fixes: 96f60dfa5819 ("trace: Use -mcount-record for dynamic ftrace")
> Signed-off-by: Greg Thelen
Looks good thanks.
Acked-by: Andi Kleen
-Andi
On Thu, Jun 07, 2018 at 12:15:12AM +0200, Jiri Olsa wrote:
> Disable stopping of the idle tick when having live cputime
> event. When the tick is disabled, the idle counts are out
> of date until next tick/update and perf cputime PMU provides
> misleading counts.
I really don't like this. This
On Thu, Jun 07, 2018 at 12:15:12AM +0200, Jiri Olsa wrote:
> Disable stopping of the idle tick when having live cputime
> event. When the tick is disabled, the idle counts are out
> of date until next tick/update and perf cputime PMU provides
> misleading counts.
I really don't like this. This
> I had some issues with IDLE counter being miscounted due to stopping
> of the idle tick. I tried to solve it in this patch (it's part of the
> patchset):
> perf/cputime: Don't stop idle tick if there's live cputime event
>
> but I'm pretty sure it's wrong and there's better solution.
At
On Thu, Jun 07, 2018 at 12:15:04AM +0200, Jiri Olsa wrote:
> Currently by default we try to match the user specified PMU
> name to all PMU units available and use them to aggregate
> all matched PMUs event counts into one 'pattern' event.
>
> While this is useful for uncore events, it screws up
On Thu, Jun 07, 2018 at 12:15:04AM +0200, Jiri Olsa wrote:
> Currently by default we try to match the user specified PMU
> name to all PMU units available and use them to aggregate
> all matched PMUs event counts into one 'pattern' event.
>
> While this is useful for uncore events, it screws up
> I had some issues with IDLE counter being miscounted due to stopping
> of the idle tick. I tried to solve it in this patch (it's part of the
> patchset):
> perf/cputime: Don't stop idle tick if there's live cputime event
>
> but I'm pretty sure it's wrong and there's better solution.
At
t; + for (cpu = cpumask_first(cpu_possible_mask), ncpu = 0;
> + cpu < num_possible_cpus() && ncpu < symnum;
> + cpu = cpumask_next(cpu, cpu_possible_mask), ncpu++)
> + ;
That is max_t(unsigned, cpumask_last(cpu_possible_mask), symnum)
Rest and ot
t; + for (cpu = cpumask_first(cpu_possible_mask), ncpu = 0;
> + cpu < num_possible_cpus() && ncpu < symnum;
> + cpu = cpumask_next(cpu, cpu_possible_mask), ncpu++)
> + ;
That is max_t(unsigned, cpumask_last(cpu_possible_mask), symnum)
Rest and ot
On Thu, May 31, 2018 at 05:08:12PM +0300, Alexey Budankov wrote:
>
> Enable complex event names containing [.:=,] symbols to be encoded into Perf
> trace using name= modifier e.g. like this:
>
> perf record -e
> cpu/name=\'OFFCORE_RESPONSE:request=DEMAND_RFO:response=L3_HIT.SNOOP_HITM\',\
>
On Thu, May 31, 2018 at 05:08:12PM +0300, Alexey Budankov wrote:
>
> Enable complex event names containing [.:=,] symbols to be encoded into Perf
> trace using name= modifier e.g. like this:
>
> perf record -e
> cpu/name=\'OFFCORE_RESPONSE:request=DEMAND_RFO:response=L3_HIT.SNOOP_HITM\',\
>
quite a few fields we report now, for example,
> LBRs,data source,weight,transaction,iregs,uregs,and etc.
>
> This patch adds these fields for perf script python.
Patches look good to me.
Reviewed-by: Andi Kleen
-Andi
quite a few fields we report now, for example,
> LBRs,data source,weight,transaction,iregs,uregs,and etc.
>
> This patch adds these fields for perf script python.
Patches look good to me.
Reviewed-by: Andi Kleen
-Andi
On Wed, May 30, 2018 at 10:20:45PM +0800, Jin Yao wrote:
> When doing pmu sampling and then running a script with
> perf script -s script.py, the process_event function gets
> dictionary with some fields from the perf ring buffer
> (like ip, sym, callchain etc).
>
> But we miss quite a few fields
On Wed, May 30, 2018 at 10:20:45PM +0800, Jin Yao wrote:
> When doing pmu sampling and then running a script with
> perf script -s script.py, the process_event function gets
> dictionary with some fields from the perf ring buffer
> (like ip, sym, callchain etc).
>
> But we miss quite a few fields
> IMHO, it is unsafe for CBOX pmu but could IMC, UPI pmus be an exception here?
> Because currently perf stat -I from IMC, UPI counters is only allowed when
> system wide monitoring is permitted and this prevents joint perf record and
> perf stat -I in cluster environments where users usually
> IMHO, it is unsafe for CBOX pmu but could IMC, UPI pmus be an exception here?
> Because currently perf stat -I from IMC, UPI counters is only allowed when
> system wide monitoring is permitted and this prevents joint perf record and
> perf stat -I in cluster environments where users usually
On Mon, May 21, 2018 at 07:13:37AM -0400, tedheadster wrote:
> Andi,
> I have a K6 regression testing system. I think my K6 is a revision
> C, but I probably can get an earlier cpu (with the bug) to test on.
>
> Do you have any specific tests you would want me to do on the affected cpu?
You
On Mon, May 21, 2018 at 07:13:37AM -0400, tedheadster wrote:
> Andi,
> I have a K6 regression testing system. I think my K6 is a revision
> C, but I probably can get an earlier cpu (with the bug) to test on.
>
> Do you have any specific tests you would want me to do on the affected cpu?
You
On Mon, May 07, 2018 at 02:30:32PM -0700, Josh Hunt wrote:
> On 05/07/2018 11:40 AM, Andi Kleen wrote:
> > On Mon, May 07, 2018 at 02:24:16PM -0400, Josh Hunt wrote:
> > > Add the ability to specify a DSO in the /tmp/perf-.map file.
> > > The DSO should be the first li
On Mon, May 07, 2018 at 02:30:32PM -0700, Josh Hunt wrote:
> On 05/07/2018 11:40 AM, Andi Kleen wrote:
> > On Mon, May 07, 2018 at 02:24:16PM -0400, Josh Hunt wrote:
> > > Add the ability to specify a DSO in the /tmp/perf-.map file.
> > > The DSO should be the first li
On Mon, May 07, 2018 at 02:24:16PM -0400, Josh Hunt wrote:
> Add the ability to specify a DSO in the /tmp/perf-.map file.
> The DSO should be the first line in the file and readable by the
> running user. If a valid DSO is found all other contents of the
> file will be ignored. This allows things
On Mon, May 07, 2018 at 02:24:16PM -0400, Josh Hunt wrote:
> Add the ability to specify a DSO in the /tmp/perf-.map file.
> The DSO should be the first line in the file and readable by the
> running user. If a valid DSO is found all other contents of the
> file will be ignored. This allows things
Jiri Olsa writes:
Please fix this quickly, PT is currently totally non functional in Linus
mainline.
-Andi
Jiri Olsa writes:
Please fix this quickly, PT is currently totally non functional in Linus
mainline.
-Andi
On Mon, Apr 30, 2018 at 10:52:56AM -0300, Arnaldo Carvalho de Melo wrote:
> Em Fri, Apr 27, 2018 at 07:52:06PM +0100, Colin King escreveu:
> > From: Colin Ian King
> >
> > Trivial fix to spelling mistakes in json text strings
>
> These files are generated from other
On Mon, Apr 30, 2018 at 10:52:56AM -0300, Arnaldo Carvalho de Melo wrote:
> Em Fri, Apr 27, 2018 at 07:52:06PM +0100, Colin King escreveu:
> > From: Colin Ian King
> >
> > Trivial fix to spelling mistakes in json text strings
>
> These files are generated from other files, so any future update
> Pekka, Andi, Stephane, do you guys see any problems with this?
Looks ok to me.
-Andi
>
> If this flies, tools/perf/Documentation/jit-interface.txt needs
> updating.
> Pekka, Andi, Stephane, do you guys see any problems with this?
Looks ok to me.
-Andi
>
> If this flies, tools/perf/Documentation/jit-interface.txt needs
> updating.
On Wed, Apr 25, 2018 at 12:02:49AM +0900, Masanari Iida wrote:
> This patch fix double words "are are".
These files are auto generated from another source. I don't think
it makes much sense to do a lot of editing on the Linux copies.
I will pass it on to the owners of the original files.
-Andi
On Wed, Apr 25, 2018 at 12:02:49AM +0900, Masanari Iida wrote:
> This patch fix double words "are are".
These files are auto generated from another source. I don't think
it makes much sense to do a lot of editing on the Linux copies.
I will pass it on to the owners of the original files.
-Andi
Commit-ID: 6a02f06edea5a5910c787fd6c49b0552e8080e5d
Gitweb: https://git.kernel.org/tip/6a02f06edea5a5910c787fd6c49b0552e8080e5d
Author: Andi Kleen <a...@linux.intel.com>
AuthorDate: Fri, 6 Apr 2018 13:38:10 -0700
Committer: Arnaldo Carvalho de Melo <a...@redhat.com>
CommitD
Commit-ID: 6a02f06edea5a5910c787fd6c49b0552e8080e5d
Gitweb: https://git.kernel.org/tip/6a02f06edea5a5910c787fd6c49b0552e8080e5d
Author: Andi Kleen
AuthorDate: Fri, 6 Apr 2018 13:38:10 -0700
Committer: Arnaldo Carvalho de Melo
CommitDate: Wed, 18 Apr 2018 15:35:49 -0300
perf hists
Commit-ID: ccbb6afe0890b09cc828373a9a5fffab40ec85df
Gitweb: https://git.kernel.org/tip/ccbb6afe0890b09cc828373a9a5fffab40ec85df
Author: Andi Kleen <a...@linux.intel.com>
AuthorDate: Fri, 6 Apr 2018 13:38:12 -0700
Committer: Arnaldo Carvalho de Melo <a...@redhat.com>
CommitD
Commit-ID: ccbb6afe0890b09cc828373a9a5fffab40ec85df
Gitweb: https://git.kernel.org/tip/ccbb6afe0890b09cc828373a9a5fffab40ec85df
Author: Andi Kleen
AuthorDate: Fri, 6 Apr 2018 13:38:12 -0700
Committer: Arnaldo Carvalho de Melo
CommitDate: Wed, 18 Apr 2018 15:35:50 -0300
perf record
Commit-ID: ec3948451e0ba317e66873b48d6cc51d701d4eb0
Gitweb: https://git.kernel.org/tip/ec3948451e0ba317e66873b48d6cc51d701d4eb0
Author: Andi Kleen <a...@linux.intel.com>
AuthorDate: Fri, 6 Apr 2018 13:38:11 -0700
Committer: Arnaldo Carvalho de Melo <a...@redhat.com>
CommitD
Commit-ID: ec3948451e0ba317e66873b48d6cc51d701d4eb0
Gitweb: https://git.kernel.org/tip/ec3948451e0ba317e66873b48d6cc51d701d4eb0
Author: Andi Kleen
AuthorDate: Fri, 6 Apr 2018 13:38:11 -0700
Committer: Arnaldo Carvalho de Melo
CommitDate: Wed, 18 Apr 2018 15:35:49 -0300
perf record
Commit-ID: a7e9eab3dbd35268c16244557a4155a2d9a641c3
Gitweb: https://git.kernel.org/tip/a7e9eab3dbd35268c16244557a4155a2d9a641c3
Author: Andi Kleen <a...@linux.intel.com>
AuthorDate: Fri, 6 Apr 2018 13:38:09 -0700
Committer: Arnaldo Carvalho de Melo <a...@redhat.com>
CommitD
Commit-ID: a7e9eab3dbd35268c16244557a4155a2d9a641c3
Gitweb: https://git.kernel.org/tip/a7e9eab3dbd35268c16244557a4155a2d9a641c3
Author: Andi Kleen
AuthorDate: Fri, 6 Apr 2018 13:38:09 -0700
Committer: Arnaldo Carvalho de Melo
CommitDate: Wed, 18 Apr 2018 15:35:48 -0300
perf mem: Allow
On Wed, Apr 18, 2018 at 05:02:02PM -0700, Linus Torvalds wrote:
> On Wed, Apr 18, 2018 at 4:26 PM, Andi Kleen <a...@linux.intel.com> wrote:
> >
> > Seems like a hack. Why can't that be stored in a per cpu variable?
>
> It *is* a percpu variable - the whole x86_tss
On Wed, Apr 18, 2018 at 05:02:02PM -0700, Linus Torvalds wrote:
> On Wed, Apr 18, 2018 at 4:26 PM, Andi Kleen wrote:
> >
> > Seems like a hack. Why can't that be stored in a per cpu variable?
>
> It *is* a percpu variable - the whole x86_tss structure is percpu.
Joerg Roedel writes:
> From: Joerg Roedel
>
> We want x86_tss.sp0 point to the entry stack later to use
> it as a trampoline stack for other kernel entry points
> besides SYSENTER.
>
> So store the task stack pointer in x86_tss.sp1, which is
> otherwise unused
Joerg Roedel writes:
> From: Joerg Roedel
>
> We want x86_tss.sp0 point to the entry stack later to use
> it as a trampoline stack for other kernel entry points
> besides SYSENTER.
>
> So store the task stack pointer in x86_tss.sp1, which is
> otherwise unused by the hardware, as Linux doesn't
> Single-stepping is only a subset of the rseq limitations addressed
> by cpu_opv. Anoher major limitation is algorithms requiring data
> migration between per-cpu data structures safely against CPU hotplug,
> and without having to change the cpu affinity mask. This is the case
And how many
> Single-stepping is only a subset of the rseq limitations addressed
> by cpu_opv. Anoher major limitation is algorithms requiring data
> migration between per-cpu data structures safely against CPU hotplug,
> and without having to change the cpu affinity mask. This is the case
And how many
> What do I miss? Or where is it that I'm misinterpreting the calculations
> that objdump did in its output?
The calculations are right, but these are still two different address modes.
You cannot just turn one silently into the other.
I think it would be ok to use the syntax in the assembler
> What do I miss? Or where is it that I'm misinterpreting the calculations
> that objdump did in its output?
The calculations are right, but these are still two different address modes.
You cannot just turn one silently into the other.
I think it would be ok to use the syntax in the assembler
On Fri, Apr 13, 2018 at 11:01:11AM -0300, Arnaldo Carvalho de Melo wrote:
> From: Arnaldo Carvalho de Melo
>
> Just like is done for 'mov' and others that can have as source or
> targets variables resolved by objdump, to make them more compact:
>
> - orb
On Fri, Apr 13, 2018 at 11:01:11AM -0300, Arnaldo Carvalho de Melo wrote:
> From: Arnaldo Carvalho de Melo
>
> Just like is done for 'mov' and others that can have as source or
> targets variables resolved by objdump, to make them more compact:
>
> - orb$0x4,0x224d71(%rip)
> Can we plan on merging just the plain rseq parts *without* this all
> first, and then see the cpu_opv thing as a "maybe future expansion"
> part.
That would be the right way to go. I doubt anybody really needs cpu_opv.
We already have other code (e.g. vgettimeofday) which cannot
be single
> Can we plan on merging just the plain rseq parts *without* this all
> first, and then see the cpu_opv thing as a "maybe future expansion"
> part.
That would be the right way to go. I doubt anybody really needs cpu_opv.
We already have other code (e.g. vgettimeofday) which cannot
be single
This patchkit fixes some random minor issues in the perf user tools
-Andi
This patchkit fixes some random minor issues in the perf user tools
-Andi
From: Andi Kleen <a...@linux.intel.com>
For perf mem report / perf mem record, pass all unknown options
through to the underlying report/record commands. This makes things
like
perf mem record -a sleep 1
work. Matches how c2c and other tools work.
Signed-off-by: Andi
From: Andi Kleen
For perf mem report / perf mem record, pass all unknown options
through to the underlying report/record commands. This makes things
like
perf mem record -a sleep 1
work. Matches how c2c and other tools work.
Signed-off-by: Andi Kleen
---
tools/perf/Documentation/perf
From: Andi Kleen <a...@linux.intel.com>
When perf record encounters an error setting up an event it suggests
to enable CONFIG_PERF_EVENTS. This is misleading because:
- Usually it is enabled (it is really hard to disable on x86)
- The problem is usually somewhere else, e.g. t
1001 - 1100 of 19746 matches
Mail list logo