From: Liu Ping Fan <qemulist@gmail.com>
To: qemu-devel@nongnu.org
Cc: Peter Maydell <peter.maydell@linaro.org>,
Jan Kiszka <jan.kiszka@siemens.com>,
Marcelo Tosatti <mtosatti@redhat.com>,
Avi Kivity <avi@redhat.com>,
Anthony Liguori <anthony@codemonkey.ws>,
Stefan Hajnoczi <stefanha@gmail.com>,
Paolo Bonzini <pbonzini@redhat.com>
Subject: [Qemu-devel] [PATCH v6 7/8] memory: introduce tls context to trace nested mmio request issue
Date: Mon, 5 Nov 2012 13:38:43 +0800 [thread overview]
Message-ID: <1352093924-17598-8-git-send-email-qemulist@gmail.com> (raw)
In-Reply-To: <1352093924-17598-1-git-send-email-qemulist@gmail.com>
From: Liu Ping Fan <pingfank@linux.vnet.ibm.com>
After breaking down big lock, nested MMIO request which not targeting
at RAM can cause deadlock issue. Supposing the scene: dev_a,b with
fine-grain locks lockA/B, then ABBA dealock issue can be triggered.
We fix this by tracing and rejecting such request.
Signed-off-by: Liu Ping Fan <pingfank@linux.vnet.ibm.com>
---
exec.c | 47 +++++++++++++++++++++++++++++++++++++++++++++++
qemu-thread.h | 7 +++++++
2 files changed, 54 insertions(+), 0 deletions(-)
diff --git a/exec.c b/exec.c
index fa34ef9..1eb920d 100644
--- a/exec.c
+++ b/exec.c
@@ -3442,6 +3442,48 @@ static bool address_space_section_lookup_ref(AddressSpace *as,
return safe_ref;
}
+typedef struct ThreadContext {
+ DispatchType dispatch_type;
+ unsigned int mmio_req_pending;
+} ThreadContext;
+
+static __thread ThreadContext thread_context = {
+ .dispatch_type = DISPATCH_INIT,
+ .mmio_req_pending = 0
+};
+
+void qemu_thread_set_dispatch_type(DispatchType type)
+{
+ thread_context.dispatch_type = type;
+}
+
+void qemu_thread_reset_dispatch_type(void)
+{
+ thread_context.dispatch_type = DISPATCH_INIT;
+}
+
+static void address_space_check_inc_req_pending(MemoryRegionSection *section)
+{
+ bool nested = false;
+
+ /* currently, only mmio out of big lock, and need this to avoid dead lock */
+ if (thread_context.dispatch_type == DISPATCH_MMIO) {
+ nested = ++thread_context.mmio_req_pending > 1 ? true : false;
+ /* To fix, will filter iommu case */
+ if (nested && !memory_region_is_ram(section->mr)) {
+ fprintf(stderr, "mmio: nested target not RAM is not support");
+ abort();
+ }
+ }
+}
+
+static void address_space_dec_req_pending(void)
+{
+ if (thread_context.dispatch_type == DISPATCH_MMIO) {
+ thread_context.mmio_req_pending--;
+ }
+}
+
void address_space_rw(AddressSpace *as, target_phys_addr_t addr, uint8_t *buf,
int len, bool is_write)
{
@@ -3462,6 +3504,8 @@ void address_space_rw(AddressSpace *as, target_phys_addr_t addr, uint8_t *buf,
qemu_mutex_lock(&as->lock);
safe_ref = memory_region_section_lookup_ref(d, page, &obj_mrs);
qemu_mutex_unlock(&as->lock);
+ address_space_check_inc_req_pending(&obj_mrs);
+
if (!safe_ref) {
qemu_mutex_lock_iothread();
qemu_mutex_lock(&as->lock);
@@ -3477,6 +3521,7 @@ void address_space_rw(AddressSpace *as, target_phys_addr_t addr, uint8_t *buf,
if (is_write) {
if (!memory_region_is_ram(section->mr)) {
target_phys_addr_t addr1;
+
addr1 = memory_region_section_addr(section, addr);
/* XXX: could force cpu_single_env to NULL to avoid
potential bugs */
@@ -3510,6 +3555,7 @@ void address_space_rw(AddressSpace *as, target_phys_addr_t addr, uint8_t *buf,
if (!(memory_region_is_ram(section->mr) ||
memory_region_is_romd(section->mr))) {
target_phys_addr_t addr1;
+
/* I/O case */
addr1 = memory_region_section_addr(section, addr);
if (l >= 4 && ((addr1 & 3) == 0)) {
@@ -3537,6 +3583,7 @@ void address_space_rw(AddressSpace *as, target_phys_addr_t addr, uint8_t *buf,
qemu_put_ram_ptr(ptr);
}
}
+ address_space_dec_req_pending();
memory_region_section_unref(&obj_mrs);
len -= l;
buf += l;
diff --git a/qemu-thread.h b/qemu-thread.h
index 05fdaaf..fc9e17b 100644
--- a/qemu-thread.h
+++ b/qemu-thread.h
@@ -7,6 +7,11 @@
typedef struct QemuMutex QemuMutex;
typedef struct QemuCond QemuCond;
typedef struct QemuThread QemuThread;
+typedef enum {
+ DISPATCH_INIT = 0,
+ DISPATCH_MMIO,
+ DISPATCH_IO,
+} DispatchType;
#ifdef _WIN32
#include "qemu-thread-win32.h"
@@ -46,4 +51,6 @@ void qemu_thread_get_self(QemuThread *thread);
bool qemu_thread_is_self(QemuThread *thread);
void qemu_thread_exit(void *retval);
+void qemu_thread_set_dispatch_type(DispatchType type);
+void qemu_thread_reset_dispatch_type(void);
#endif
--
1.7.4.4
next prev parent reply other threads:[~2012-11-05 5:39 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-11-05 5:38 [Qemu-devel] [PATCH v6 0/8] push mmio dispatch out of big lock Liu Ping Fan
2012-11-05 5:38 ` [Qemu-devel] [PATCH v6 1/8] atomic: introduce atomic operations Liu Ping Fan
2012-11-12 9:54 ` Paolo Bonzini
2012-11-13 6:48 ` liu ping fan
2012-11-13 10:11 ` Paolo Bonzini
2012-11-14 9:38 ` liu ping fan
2012-11-14 9:47 ` Paolo Bonzini
2012-11-15 7:47 ` liu ping fan
2012-11-15 11:24 ` Paolo Bonzini
2012-11-16 0:03 ` Richard Henderson
2012-11-21 5:58 ` liu ping fan
2012-11-18 10:04 ` Avi Kivity
2012-11-21 5:57 ` liu ping fan
2012-11-13 10:11 ` Paolo Bonzini
2012-11-05 5:38 ` [Qemu-devel] [PATCH v6 2/8] qom: apply atomic on object's refcount Liu Ping Fan
2012-11-05 5:38 ` [Qemu-devel] [PATCH v6 3/8] hotplug: introduce qdev_unplug_complete() to remove device from views Liu Ping Fan
2012-11-12 9:27 ` Paolo Bonzini
2012-11-13 6:12 ` liu ping fan
2012-11-05 5:38 ` [Qemu-devel] [PATCH v6 4/8] pci: remove pci device from mem view when unplug Liu Ping Fan
2012-11-05 5:38 ` [Qemu-devel] [PATCH v6 5/8] memory: introduce local lock for address space Liu Ping Fan
2012-11-05 5:38 ` [Qemu-devel] [PATCH v6 6/8] memory: make mmio dispatch able to be out of biglock Liu Ping Fan
2012-11-05 6:45 ` Jan Kiszka
2012-11-05 5:38 ` Liu Ping Fan [this message]
2012-11-05 6:57 ` [Qemu-devel] [PATCH v6 7/8] memory: introduce tls context to trace nested mmio request issue Jan Kiszka
2012-11-05 5:38 ` [Qemu-devel] [PATCH v6 8/8] vcpu: push mmio dispatcher out of big lock Liu Ping Fan
2012-11-05 7:00 ` [Qemu-devel] [PATCH v6 0/8] push mmio dispatch " Jan Kiszka
2012-11-09 6:23 ` liu ping fan
2012-11-09 8:15 ` Jan Kiszka
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=1352093924-17598-8-git-send-email-qemulist@gmail.com \
--to=qemulist@gmail.com \
--cc=anthony@codemonkey.ws \
--cc=avi@redhat.com \
--cc=jan.kiszka@siemens.com \
--cc=mtosatti@redhat.com \
--cc=pbonzini@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@gmail.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).