From: "Alex Bennée" <alex.bennee@linaro.org>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: mttcg@listserver.greensocs.com, mark.burton@greensocs.com,
a.rigo@virtualopensystems.com, qemu-devel@nongnu.org,
guillaume.delbergue@greensocs.com, fred.konrad@greensocs.com
Subject: Re: [Qemu-devel] [RFC PATCH V7 13/19] add a callback when tb_invalidate is called.
Date: Mon, 10 Aug 2015 19:41:13 +0100 [thread overview]
Message-ID: <877fp3gf5i.fsf@linaro.org> (raw)
In-Reply-To: <55C8D6D2.4030908@redhat.com>
Paolo Bonzini <pbonzini@redhat.com> writes:
> On 10/08/2015 17:27, fred.konrad@greensocs.com wrote:
>> From: KONRAD Frederic <fred.konrad@greensocs.com>
>>
>> Instead of doing the jump cache invalidation directly in tb_invalidate delay it
>> after the exit so we don't have an other CPU trying to execute the code being
>> invalidated.
>>
>> Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
>> ---
>> translate-all.c | 61 +++++++++++++++++++++++++++++++++++++++++++++++++++++++--
>> 1 file changed, 59 insertions(+), 2 deletions(-)
>
> If you take the easy way and avoid the optimizations in patch 7, this is
> not necessary: tb_find_fast and tb_add_jump are only called from within
> tb_lock, so all of tb_jmp_cache/jmp_first/jmp_next are protected by tb_lock.
>
> Let's get everything in and then optimize; the order should be:
>
> - Alvise's LL/SC implementation
>
> - conversion of atomics to LL/SC for all front-ends
>
> - the main MTTCG series, reusing the locking already in-place for
> user-mode emulation (with some audit...)
- including dropping the cmpxchg fix and including Alvise's MTTCG aware patches
that build on top of LL/SC work.
>
> - any further push-downs of tb_lock
>
> Paolo
>
>> diff --git a/translate-all.c b/translate-all.c
>> index 954c67a..fc5162a 100644
>> --- a/translate-all.c
>> +++ b/translate-all.c
>> @@ -62,6 +62,7 @@
>> #include "translate-all.h"
>> #include "qemu/bitmap.h"
>> #include "qemu/timer.h"
>> +#include "sysemu/cpus.h"
>>
>> //#define DEBUG_TB_INVALIDATE
>> //#define DEBUG_FLUSH
>> @@ -967,14 +968,58 @@ static inline void tb_reset_jump(TranslationBlock *tb, int n)
>> tb_set_jmp_target(tb, n, (uintptr_t)(tb->tc_ptr + tb->tb_next_offset[n]));
>> }
>>
>> +struct CPUDiscardTBParams {
>> + CPUState *cpu;
>> + TranslationBlock *tb;
>> +};
>> +
>> +static void cpu_discard_tb_from_jmp_cache(void *opaque)
>> +{
>> + unsigned int h;
>> + struct CPUDiscardTBParams *params = opaque;
>> +
>> + h = tb_jmp_cache_hash_func(params->tb->pc);
>> + if (params->cpu->tb_jmp_cache[h] == params->tb) {
>> + params->cpu->tb_jmp_cache[h] = NULL;
>> + }
>> +
>> + g_free(opaque);
>> +}
>> +
>> +static void tb_invalidate_jmp_remove(void *opaque)
>> +{
>> + TranslationBlock *tb = opaque;
>> + TranslationBlock *tb1, *tb2;
>> + unsigned int n1;
>> +
>> + /* suppress this TB from the two jump lists */
>> + tb_jmp_remove(tb, 0);
>> + tb_jmp_remove(tb, 1);
>> +
>> + /* suppress any remaining jumps to this TB */
>> + tb1 = tb->jmp_first;
>> + for (;;) {
>> + n1 = (uintptr_t)tb1 & 3;
>> + if (n1 == 2) {
>> + break;
>> + }
>> + tb1 = (TranslationBlock *)((uintptr_t)tb1 & ~3);
>> + tb2 = tb1->jmp_next[n1];
>> + tb_reset_jump(tb1, n1);
>> + tb1->jmp_next[n1] = NULL;
>> + tb1 = tb2;
>> + }
>> + tb->jmp_first = (TranslationBlock *)((uintptr_t)tb | 2); /* fail safe */
>> +}
>> +
>> /* invalidate one TB */
>> void tb_phys_invalidate(TranslationBlock *tb, tb_page_addr_t page_addr)
>> {
>> CPUState *cpu;
>> PageDesc *p;
>> - unsigned int h, n1;
>> + unsigned int h;
>> tb_page_addr_t phys_pc;
>> - TranslationBlock *tb1, *tb2;
>> + struct CPUDiscardTBParams *params;
>>
>> tb_lock();
>>
>> @@ -997,6 +1042,9 @@ void tb_phys_invalidate(TranslationBlock *tb, tb_page_addr_t page_addr)
>>
>> tcg_ctx.tb_ctx.tb_invalidated_flag = 1;
>>
>> +#if 0 /*MTTCG*/
>> + TranslationBlock *tb1, *tb2;
>> + unsigned int n1;
>> /* remove the TB from the hash list */
>> h = tb_jmp_cache_hash_func(tb->pc);
>> CPU_FOREACH(cpu) {
>> @@ -1023,6 +1071,15 @@ void tb_phys_invalidate(TranslationBlock *tb, tb_page_addr_t page_addr)
>> tb1 = tb2;
>> }
>> tb->jmp_first = (TranslationBlock *)((uintptr_t)tb | 2); /* fail safe */
>> +#else
>> + CPU_FOREACH(cpu) {
>> + params = g_malloc(sizeof(struct CPUDiscardTBParams));
>> + params->cpu = cpu;
>> + params->tb = tb;
>> + async_run_on_cpu(cpu, cpu_discard_tb_from_jmp_cache, params);
>> + }
>> + async_run_safe_work_on_cpu(first_cpu, tb_invalidate_jmp_remove, tb);
>> +#endif /* MTTCG */
>>
>> tcg_ctx.tb_ctx.tb_phys_invalidate_count++;
>> tb_unlock();
>>
--
Alex Bennée
next prev parent reply other threads:[~2015-08-10 18:41 UTC|newest]
Thread overview: 81+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-08-10 15:26 [Qemu-devel] [RFC PATCH V7 00/19] Multithread TCG fred.konrad
2015-08-10 15:26 ` [Qemu-devel] [RFC PATCH V7 01/19] cpus: protect queued_work_* with work_mutex fred.konrad
2015-08-10 15:59 ` Paolo Bonzini
2015-08-10 16:04 ` Frederic Konrad
2015-08-10 16:06 ` Paolo Bonzini
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 02/19] cpus: add tcg_exec_flag fred.konrad
2015-08-11 10:53 ` Paolo Bonzini
2015-08-11 11:11 ` Frederic Konrad
2015-08-11 12:57 ` Paolo Bonzini
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 03/19] cpus: introduce async_run_safe_work_on_cpu fred.konrad
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 04/19] replace spinlock by QemuMutex fred.konrad
2015-08-10 16:09 ` Paolo Bonzini
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 05/19] remove unused spinlock fred.konrad
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 06/19] add support for spin lock on POSIX systems exclusively fred.konrad
2015-08-10 16:10 ` Paolo Bonzini
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 07/19] protect TBContext with tb_lock fred.konrad
2015-08-10 16:36 ` Paolo Bonzini
2015-08-10 16:50 ` Paolo Bonzini
2015-08-10 18:39 ` Alex Bennée
2015-08-11 8:31 ` Paolo Bonzini
2015-08-11 6:46 ` Frederic Konrad
2015-08-11 8:34 ` Paolo Bonzini
2015-08-11 9:21 ` Peter Maydell
2015-08-11 9:59 ` Paolo Bonzini
2015-08-12 17:45 ` Frederic Konrad
2015-08-12 18:20 ` Alex Bennée
2015-08-12 18:22 ` Paolo Bonzini
2015-08-14 8:38 ` Frederic Konrad
2015-08-15 0:04 ` Paolo Bonzini
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 08/19] tcg: remove tcg_halt_cond global variable fred.konrad
2015-08-10 16:12 ` Paolo Bonzini
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 09/19] Drop global lock during TCG code execution fred.konrad
2015-08-10 16:15 ` Paolo Bonzini
2015-08-11 6:55 ` Frederic Konrad
2015-08-11 20:12 ` Alex Bennée
2015-08-11 21:34 ` Frederic Konrad
2015-08-12 9:58 ` Paolo Bonzini
2015-08-12 12:32 ` Frederic Konrad
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 10/19] cpu: remove exit_request global fred.konrad
2015-08-10 15:51 ` Paolo Bonzini
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 11/19] tcg: switch on multithread fred.konrad
2015-08-13 11:17 ` Paolo Bonzini
2015-08-13 14:41 ` Frederic Konrad
2015-08-13 14:58 ` Paolo Bonzini
2015-08-13 15:18 ` Frederic Konrad
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 12/19] Use atomic cmpxchg to atomically check the exclusive value in a STREX fred.konrad
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 13/19] add a callback when tb_invalidate is called fred.konrad
2015-08-10 16:52 ` Paolo Bonzini
2015-08-10 18:41 ` Alex Bennée [this message]
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 14/19] cpu: introduce tlb_flush*_all fred.konrad
2015-08-10 15:54 ` Paolo Bonzini
2015-08-10 16:00 ` Peter Maydell
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 15/19] arm: use tlb_flush*_all fred.konrad
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 16/19] translate-all: introduces tb_flush_safe fred.konrad
2015-08-10 16:26 ` Paolo Bonzini
2015-08-12 14:09 ` Paolo Bonzini
2015-08-12 14:11 ` Frederic Konrad
2015-08-12 14:14 ` Paolo Bonzini
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 17/19] translate-all: (wip) use tb_flush_safe when we can't alloc more tb fred.konrad
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 18/19] mttcg: signal the associated cpu anyway fred.konrad
2015-08-10 15:51 ` Paolo Bonzini
2015-08-10 15:27 ` [Qemu-devel] [RFC PATCH V7 19/19] target-arm/psci.c: wake up sleeping CPUs (MTTCG) fred.konrad
2015-08-10 16:41 ` Paolo Bonzini
2015-08-10 18:38 ` Alex Bennée
2015-08-10 18:34 ` [Qemu-devel] [RFC PATCH V7 00/19] Multithread TCG Alex Bennée
2015-08-10 23:02 ` Frederic Konrad
2015-08-11 6:15 ` Benjamin Herrenschmidt
2015-08-11 6:27 ` Frederic Konrad
2015-10-07 12:46 ` Claudio Fontana
2015-10-07 14:52 ` Frederic Konrad
2015-10-21 15:09 ` Claudio Fontana
2015-08-11 7:54 ` Alex Bennée
2015-08-11 9:22 ` Benjamin Herrenschmidt
2015-08-11 9:29 ` Peter Maydell
2015-08-11 10:09 ` Benjamin Herrenschmidt
2015-08-11 19:22 ` Alex Bennée
2015-08-11 12:45 ` Paolo Bonzini
2015-08-11 13:59 ` Frederic Konrad
2015-08-11 14:10 ` Paolo Bonzini
2015-08-12 15:19 ` Frederic Konrad
2015-08-12 15:39 ` Paolo Bonzini
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=877fp3gf5i.fsf@linaro.org \
--to=alex.bennee@linaro.org \
--cc=a.rigo@virtualopensystems.com \
--cc=fred.konrad@greensocs.com \
--cc=guillaume.delbergue@greensocs.com \
--cc=mark.burton@greensocs.com \
--cc=mttcg@listserver.greensocs.com \
--cc=pbonzini@redhat.com \
--cc=qemu-devel@nongnu.org \
/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).