COLO: Fix memory leak in packet_enqueue()
v2 cover COLO: Fix memory leak in packet_enqueue()
The patch is to fix the memory leak in packet_enqueue(). The allocated "pkt" needs to be freed if the colo compare primary or secondary queue is too big to insert.
Reproduce steps:
(1) Setup PVM and SVM both with NIC e1000 by the steps descripted in the wiki qemu/COLO
(2) Run "iperf3 -s" in PVM
(3) Run "iperfs -c <PVM-IP> -t 7200"
The memory usage of qemu-system-x86_64 increases as the PVM's QMP shows "qemu-system-x86_64: colo compare secondary queue size too big, drop packet".
Please review, thanks.
V2:
- Fix incorrect patch format
Derek Su (1):
net/colo-compare.c: Fix memory leak in packet_enqueue()
net/colo-compare.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-)
Signature · 1 line
2.17.1
PATCH v2 1/1 net/colo-compare.c: Fix memory leak in packet_enqueue()
The patch is to fix the "pkt" memory leak in packet_enqueue(). The allocated "pkt" needs to be freed if the colo compare primary or secondary queue is too big.
--- net/colo-compare.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-)
net/colo-compare.c +15 -8
@@ -120,6 +120,10 @@ enum { SECONDARY_IN, }; +static const char *colo_mode[] = { + [PRIMARY_IN ] ="primary" , + [SECONDARY_IN ] ="secondary" , +}; static int compare_chr_send(CompareState *s, const uint8_t *buf,
@@ -215,6 +219,7 @@ static int packet_enqueue(CompareState *s, int mode, Connection **con)ConnectionKey key;Packet *pkt =NULL ;Connection *conn; +int ret;if (mode ==PRIMARY_IN ) { pkt = packet_new(s->pri_rs.buf,
@@ -243,16 +248,18 @@ static int packet_enqueue(CompareState *s, int mode, Connection **con) }if (mode ==PRIMARY_IN ) { -if (!colo_insert_packet (&conn->primary_list, pkt, &conn->pack)) { -error_report ("colo compare primary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->primary_list, pkt, &conn->pack); }else { -if (!colo_insert_packet (&conn->secondary_list, pkt, &conn->sack)) { -error_report ("colo compare secondary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->secondary_list, pkt, &conn->sack); } + +if (!ret) { +error_report ("colo compare %s queue size too big," +"drop packet" , colo_mode[mode]); +packet_destroy (pkt,NULL ); + pkt =NULL ; + } + *con = conn;return 0 ;
Signature · 1 line
2.17.1
RE: [PATCH v2 1/1] net/colo-compare.c: Fix memory leak in packet_enqueue()
Quoted 11 lines
-----Original Message-----
From: Derek Su <[email protected]>
Sent: Monday, March 23, 2020 1:48 AM
To: [email protected]
Cc: Zhang, Chen <[email protected]>; [email protected]; [email protected]; [email protected]
Subject: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()
Quoted 3 lines
The patch is to fix the "pkt" memory leak in
packet_enqueue(). The allocated "pkt" needs to be freed if the colo compare primary or secondary queue is too big.
Hi Derek,
Thank you for the patch.
I re-think this issue in a big view, looks just free the pkg is not enough in this situation. The root cause is network is too busy to compare, So, better choice is notify COLO frame to do a checkpoint and clean up all the network queue. This work maybe decrease COLO network performance but seams better than drop lots of pkg.
Thanks
Zhang Chen
Quoted 60 lines
Signed-off-by Derek Su <[email protected]>--- net/colo-compare.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-)net/colo-compare.c index renamed from net/colo-compare.c +0
-0
7ee17f2cf8..cdd87b2aa8100644net/colo-compare.c +4
-0@@ -120,6 +120,10 @@ enum { SECONDARY_IN, }; +static const char *colo_mode[] = { + [PRIMARY_IN ] ="primary" , + [SECONDARY_IN ] ="secondary" , +};static int compare_chr_send (CompareState *s, const uint8_t *buf, @@ -215,6 +219,7 @@ static int
packet_enqueue(CompareState *s, int mode, Connection **con)ConnectionKey key; Packet *pkt = NULL; Connection *conn;+ int ret;
if (mode == PRIMARY_IN) { pkt = packet_new(s->pri_rs.buf, @@ -243,16 +248,18 @@ static int packet_enqueue(CompareState *s, int mode, Connection **con) } if (mode == PRIMARY_IN) {net/colo-compare.c +12
-9-if (!colo_insert_packet (&conn->primary_list, pkt, &conn->pack)) { -error_report ("colo compare primary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->primary_list, pkt, + &conn->pack); }else { -if (!colo_insert_packet (&conn->secondary_list, pkt, &conn->sack)) { -error_report ("colo compare secondary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->secondary_list, pkt, + &conn->sack); } + +if (!ret) { +error_report ("colo compare %s queue size too big," +"drop packet" , colo_mode[mode]); +packet_destroy (pkt,NULL ); + pkt =NULL ; + } + *con = conn;return 0 ; --2.17.1
Re: [PATCH v2 1/1] net/colo-compare.c: Fix memory leak in packet_enqueue()
Zhang, Chen <[email protected]> 於 2020年3月24日 週二 上午3:24寫道:
Quoted 18 lines
Quoted 12 lines
-----Original Message-----
From: Derek Su <[email protected]>
Sent: Monday, March 23, 2020 1:48 AM
To: [email protected]
Cc: Zhang, Chen <[email protected]>; [email protected]; [email protected]; [email protected]
Subject: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Quoted 3 lines
The patch is to fix the "pkt" memory leak in
packet_enqueue(). The allocated "pkt" needs to be freed if the colo compare primary or secondary queue is too big.Hi Derek,
Quoted 9 lines
Thank you for the patch.
I re-think this issue in a big view, looks just free the pkg is not enough in this situation. The root cause is network is too busy to compare, So, better choice is notify COLO frame to do a checkpoint and clean up all the network queue. This work maybe decrease COLO network performance but seams better than drop lots of pkg.Thanks
Zhang Chen
Hello, Zhang
Got it.
What is the concern of the massive "drop packets"?
Does the behavior make the COLO do checkpoint periodically (~20 seconds) instead of doing immediate checkpoint when encountering different response packets?
It seems that frequent checkpoints caused by the full queue (busy network) instead of different
response packets may harm the high speed network (10 Gbps or higher) performance dramatically.
Thanks
Derek
Quoted 61 lines
Quoted 60 lines
Signed-off-by Derek Su <[email protected]>--- net/colo-compare.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-)net/colo-compare.c index renamed from net/colo-compare.c +0
-0
7ee17f2cf8..cdd87b2aa8100644net/colo-compare.c +4
-0@@ -120,6 +120,10 @@ enum { SECONDARY_IN, }; +static const char *colo_mode[] = { + [PRIMARY_IN ] ="primary" , + [SECONDARY_IN ] ="secondary" , +};static int compare_chr_send (CompareState *s, const uint8_t *buf, @@ -215,6 +219,7 @@ static int
packet_enqueue(CompareState *s, int mode, Connection **con)ConnectionKey key; Packet *pkt = NULL; Connection *conn;+ int ret;
if (mode == PRIMARY_IN) { pkt = packet_new(s->pri_rs.buf, @@ -243,16 +248,18 @@ static int packet_enqueue(CompareState *s, int mode, Connection **con) } if (mode == PRIMARY_IN) {net/colo-compare.c +12
-9-if (!colo_insert_packet (&conn->primary_list, pkt, &conn->pack)) { -error_report ("colo compare primary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->primary_list, pkt, + &conn->pack); }else { -if (!colo_insert_packet (&conn->secondary_list, pkt, &conn->sack)) { -error_report ("colo compare secondary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->secondary_list, pkt, + &conn->sack); } + +if (!ret) { +error_report ("colo compare %s queue size too big," +"drop packet" , colo_mode[mode]); +packet_destroy (pkt,NULL ); + pkt =NULL ; + } + *con = conn;return 0 ; --2.17.1
RE: [PATCH v2 1/1] net/colo-compare.c: Fix memory leak in packet_enqueue()
Quoted 43 lines
-----Original Message-----
From: Jing-Wei Su <[email protected]>
Sent: Tuesday, March 24, 2020 10:47 AM
To: Zhang, Chen <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Zhang, Chen <[email protected]> 於 2020年3月24日 週二 上午3:24 寫道:
Quoted 12 lines
Quoted 12 lines
-----Original Message-----
From: Derek Su <[email protected]>
Sent: Monday, March 23, 2020 1:48 AM
To: [email protected]
Cc: Zhang, Chen <[email protected]>; [email protected]; [email protected]; [email protected]
Subject: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Quoted 8 lines
Quoted 3 lines
The patch is to fix the "pkt" memory leak in
packet_enqueue(). The allocated "pkt" needs to be freed if the colo compare primary or secondary queue is too big.Hi Derek,
Thank you for the patch.
I re-think this issue in a big view, looks just free the pkg is not enough inthis situation.
Quoted 3 lines
The root cause is network is too busy to compare, So, better choice is notify COLO frame to do a checkpoint and clean up all the network queue. This work maybe decrease COLO network performance but seams
better than drop lots of pkg.
Quoted 4 lines
Thanks
Zhang Chen
Quoted 7 lines
Hello, Zhang
Got it.
What is the concern of the massive "drop packets"?
Does the behavior make the COLO do checkpoint periodically (~20 seconds) instead of doing immediate checkpoint when encountering different response packets?
The concern of the "drop packets" is guest will lose network connection with most of real clients until next periodic force checkpoint. COLO designed for dynamic control checkpoint, so I think do a checkpoint here will help guest supply service faster.
Quoted 5 lines
It seems that frequent checkpoints caused by the full queue (busy network) instead of different
response packets may harm the high speed network (10 Gbps or higher) performance dramatically.
Yes, maybe I can send a patch to make user adjust queue size depend on it's own environment. But with larger queue size, colo-compare will spend much time to do compare packet when network Is real busy status.
Thanks
Zhang Chen
Quoted 67 lines
Thanks
DerekQuoted 14 lines
Quoted 14 lines
Signed-off-by Derek Su <[email protected]>--- net/colo-compare.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-)net/colo-compare.c index renamed from net/colo-compare.c +0
-0
7ee17f2cf8..cdd87b2aa8100644net/colo-compare.c +0
-0@@ -120,6 +120,10 @@ enum { SECONDARY_IN, };Quoted 8 lines
Quoted 8 lines
net/colo-compare.c +4
-0+static const char *colo_mode[] = { + [PRIMARY_IN ] ="primary" , + [SECONDARY_IN ] ="secondary" , +}; static int compare_chr_send(CompareState *s, const uint8_t *buf, @@ -215,6 +219,7 @@static int packet_enqueue(CompareState *s, int mode, Connection
**con)
Quoted 8 lines
Quoted 8 lines
ConnectionKey key; Packet *pkt = NULL; Connection *conn;+ int ret;
if (mode == PRIMARY_IN) { pkt = packet_new(s->pri_rs.buf, @@ -243,16 +248,18 @@static int packet_enqueue(CompareState *s, int mode, Connection
**con)
Quoted 31 lines
Quoted 30 lines
} if (mode == PRIMARY_IN) {net/colo-compare.c +12
-9-if (!colo_insert_packet (&conn->primary_list, pkt, &conn->pack)) { -error_report ("colo compare primary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->primary_list, pkt, + &conn->pack); }else { -if (!colo_insert_packet (&conn->secondary_list, pkt, &conn->sack)) { -error_report ("colo compare secondary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->secondary_list, pkt, + &conn->sack); } + +if (!ret) { +error_report ("colo compare %s queue size too big," +"drop packet" , colo_mode[mode]); +packet_destroy (pkt,NULL ); + pkt =NULL ; + } + *con = conn;return 0 ; --2.17.1
Re: [PATCH v2 1/1] net/colo-compare.c: Fix memory leak in packet_enqueue()
Zhang, Chen <[email protected]> 於 2020年3月25日 週三 上午9:37寫道:
Quoted 52 lines
Quoted 44 lines
-----Original Message-----
From: Jing-Wei Su <[email protected]>
Sent: Tuesday, March 24, 2020 10:47 AM
To: Zhang, Chen <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Zhang, Chen <[email protected]> 於 2020年3月24日 週二 上午3:24 寫道:
Quoted 12 lines
Quoted 12 lines
-----Original Message-----
From: Derek Su <[email protected]>
Sent: Monday, March 23, 2020 1:48 AM
To: [email protected]
Cc: Zhang, Chen <[email protected]>; [email protected]; [email protected]; [email protected]
Subject: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Quoted 8 lines
Quoted 3 lines
The patch is to fix the "pkt" memory leak in
packet_enqueue(). The allocated "pkt" needs to be freed if the colo compare primary or secondary queue is too big.Hi Derek,
Thank you for the patch.
I re-think this issue in a big view, looks just free the pkg is not enough inthis situation.
Quoted 3 lines
The root cause is network is too busy to compare, So, better choice is notify COLO frame to do a checkpoint and clean up all the network queue. This work maybe decrease COLO network performance but seams
better than drop lots of pkg.
Quoted 4 lines
Thanks
Zhang ChenQuoted 7 lines
Hello, Zhang
Got it.
What is the concern of the massive "drop packets"?
Does the behavior make the COLO do checkpoint periodically (~20 seconds) instead of doing immediate checkpoint when encountering different response packets?
Quoted 4 lines
The concern of the "drop packets" is guest will lose network connection with most of real clients until next periodic force checkpoint. COLO designed for dynamic control checkpoint, so I think do a checkpoint here will help guest supply service faster.
I see.
I'll update the patch with your suggestion later.
Quoted 9 lines
Quoted 5 lines
It seems that frequent checkpoints caused by the full queue (busy network) instead of different
response packets may harm the high speed network (10 Gbps or higher) performance dramatically.Yes, maybe I can send a patch to make user adjust queue size depend on it's own environment. But with larger queue size, colo-compare will spend much time to do compare packet when network Is real busy status.
Thank you. The user-configurable queue size will be very helpful.
Thanks.
Derek Su
Quoted 71 lines
Thanks
Zhang ChenQuoted 67 lines
Thanks
DerekQuoted 14 lines
Quoted 14 lines
Signed-off-by Derek Su <[email protected]>--- net/colo-compare.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-)net/colo-compare.c index renamed from net/colo-compare.c +0
-0
7ee17f2cf8..cdd87b2aa8100644net/colo-compare.c +0
-0@@ -120,6 +120,10 @@ enum { SECONDARY_IN, };Quoted 8 lines
Quoted 8 lines
net/colo-compare.c +4
-0+static const char *colo_mode[] = { + [PRIMARY_IN ] ="primary" , + [SECONDARY_IN ] ="secondary" , +}; static int compare_chr_send(CompareState *s, const uint8_t *buf, @@ -215,6 +219,7 @@static int packet_enqueue(CompareState *s, int mode, Connection
**con)
Quoted 8 lines
Quoted 8 lines
ConnectionKey key; Packet *pkt = NULL; Connection *conn;+ int ret;
if (mode == PRIMARY_IN) { pkt = packet_new(s->pri_rs.buf, @@ -243,16 +248,18 @@static int packet_enqueue(CompareState *s, int mode, Connection
**con)
Quoted 31 lines
Quoted 30 lines
} if (mode == PRIMARY_IN) {net/colo-compare.c +12
-9-if (!colo_insert_packet (&conn->primary_list, pkt, &conn->pack)) { -error_report ("colo compare primary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->primary_list, pkt, + &conn->pack); }else { -if (!colo_insert_packet (&conn->secondary_list, pkt, &conn->sack)) { -error_report ("colo compare secondary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->secondary_list, pkt, + &conn->sack); } + +if (!ret) { +error_report ("colo compare %s queue size too big," +"drop packet" , colo_mode[mode]); +packet_destroy (pkt,NULL ); + pkt =NULL ; + } + *con = conn;return 0 ; --2.17.1
Re: [PATCH v2 1/1] net/colo-compare.c: Fix memory leak in packet_enqueue()
Jing-Wei Su <[email protected]> 於 2020年3月25日 週三 上午10:05寫道:
Quoted 54 lines
Zhang, Chen <[email protected]> 於 2020年3月25日 週三 上午9:37寫道:
Quoted 52 lines
Quoted 51 lines
-----Original Message-----
From: Jing-Wei Su <[email protected]>
Sent: Tuesday, March 24, 2020 10:47 AM
To: Zhang, Chen <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Zhang, Chen <[email protected]> 於 2020年3月24日 週二 上午3:24 寫道:
Quoted 12 lines
Quoted 12 lines
-----Original Message-----
From: Derek Su <[email protected]>
Sent: Monday, March 23, 2020 1:48 AM
To: [email protected]
Cc: Zhang, Chen <[email protected]>; [email protected]; [email protected]; [email protected]
Subject: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Quoted 8 lines
Quoted 3 lines
The patch is to fix the "pkt" memory leak in
packet_enqueue(). The allocated "pkt" needs to be freed if the colo compare primary or secondary queue is too big.Hi Derek,
Thank you for the patch.
I re-think this issue in a big view, looks just free the pkg is not enough inthis situation.
Quoted 3 lines
The root cause is network is too busy to compare, So, better choice is notify COLO frame to do a checkpoint and clean up all the network queue. This work maybe decrease COLO network performance but seams
better than drop lots of pkg.
Quoted 4 lines
Thanks
Zhang ChenHello, Zhang
Got it.
What is the concern of the massive "drop packets"?
Does the behavior make the COLO do checkpoint periodically (~20 seconds) instead of doing immediate checkpoint when encountering different response packets?
Quoted 8 lines
Quoted 4 lines
The concern of the "drop packets" is guest will lose network connection with most of real clients until next periodic force checkpoint. COLO designed for dynamic control checkpoint, so I think do a checkpoint here will help guest supply service faster.
I see.
I'll update the patch with your suggestion later.
Hi, Zhang
Here is the idea and pseudo code to handle the "drop packet".
```
ret = packet_enqueue
(1) ret == 0
compare connection
(2) ret == -1
send packet
(3) ret == queue insertion fail
do checkpoint
send all queued primary packets
remove all queued secondary packets
```Do you have any comment for the handling?
Thanks
Derek
Quoted 86 lines
Quoted 9 lines
Quoted 5 lines
It seems that frequent checkpoints caused by the full queue (busy network) instead of different
response packets may harm the high speed network (10 Gbps or higher) performance dramatically.Yes, maybe I can send a patch to make user adjust queue size depend on it's own environment. But with larger queue size, colo-compare will spend much time to do compare packet when network Is real busy status.
Thank you. The user-configurable queue size will be very helpful.
Thanks.
Derek SuQuoted 71 lines
Thanks
Zhang ChenQuoted 67 lines
Thanks
DerekQuoted 14 lines
Quoted 14 lines
Signed-off-by Derek Su <[email protected]>--- net/colo-compare.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-)net/colo-compare.c index renamed from net/colo-compare.c +0
-0
7ee17f2cf8..cdd87b2aa8100644net/colo-compare.c +0
-0@@ -120,6 +120,10 @@ enum { SECONDARY_IN, };Quoted 8 lines
Quoted 8 lines
net/colo-compare.c +4
-0+static const char *colo_mode[] = { + [PRIMARY_IN ] ="primary" , + [SECONDARY_IN ] ="secondary" , +}; static int compare_chr_send(CompareState *s, const uint8_t *buf, @@ -215,6 +219,7 @@static int packet_enqueue(CompareState *s, int mode, Connection
**con)
Quoted 8 lines
Quoted 8 lines
ConnectionKey key; Packet *pkt = NULL; Connection *conn;+ int ret;
if (mode == PRIMARY_IN) { pkt = packet_new(s->pri_rs.buf, @@ -243,16 +248,18 @@static int packet_enqueue(CompareState *s, int mode, Connection
**con)
Quoted 31 lines
Quoted 30 lines
} if (mode == PRIMARY_IN) {net/colo-compare.c +12
-9-if (!colo_insert_packet (&conn->primary_list, pkt, &conn->pack)) { -error_report ("colo compare primary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->primary_list, pkt, + &conn->pack); }else { -if (!colo_insert_packet (&conn->secondary_list, pkt, &conn->sack)) { -error_report ("colo compare secondary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->secondary_list, pkt, + &conn->sack); } + +if (!ret) { +error_report ("colo compare %s queue size too big," +"drop packet" , colo_mode[mode]); +packet_destroy (pkt,NULL ); + pkt =NULL ; + } + *con = conn;return 0 ; --2.17.1
RE: [PATCH v2 1/1] net/colo-compare.c: Fix memory leak in packet_enqueue()
Quoted 96 lines
-----Original Message-----
From: Derek Su <[email protected]>
Sent: Wednesday, March 25, 2020 12:17 PM
To: Zhang, Chen <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Jing-Wei Su <[email protected]> 於 2020年3月25日 週三 上午10:05 寫道:
Quoted 2 lines
Zhang, Chen <[email protected]> 於 2020年3月25日 週三 上午
9:37寫道:
Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
-----Original Message-----
From: Jing-Wei Su <[email protected]>
Sent: Tuesday, March 24, 2020 10:47 AM
To: Zhang, Chen <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Quoted 1 line
Quoted 1 line
Quoted 1 line
Zhang, Chen <[email protected]> 於 2020年3月24日 週二 上午
3:24
Quoted 42 lines
Quoted 42 lines
Quoted 41 lines
寫道:
Quoted 17 lines
Quoted 13 lines
-----Original Message-----
From: Derek Su <[email protected]>
Sent: Monday, March 23, 2020 1:48 AM
To: [email protected]
Cc: Zhang, Chen <[email protected]>; [email protected]; [email protected]; [email protected]
Subject: [PATCH v2 1/1]net/colo-compare.c: Fix memory leak inpacket_enqueue()Quoted 3 lines
The patch is to fix the "pkt" memory leak in
packet_enqueue(). The allocated "pkt" needs to be freed if the colo compare primary or secondary queue is too big.Quoted 5 lines
Hi Derek,
Thank you for the patch.
I re-think this issue in a big view, looks just free the pkg is not enough inthis situation.
Quoted 4 lines
The root cause is network is too busy to compare, So, better choice is notify COLO frame to do a checkpoint and clean up all the network queue. This work maybe decrease COLO network performance but seams
better than drop lots of pkg.
Quoted 4 lines
Thanks
Zhang ChenHello, Zhang
Got it.
What is the concern of the massive "drop packets"? Does the behavior make the COLO do checkpoint periodically (~20 seconds) instead of doing immediate checkpoint when encountering different response packets?Quoted 3 lines
Quoted 3 lines
The concern of the "drop packets" is guest will lose network connection with most of real clients until next periodic force checkpoint. COLO designed for dynamic control checkpoint, so I think do
a checkpoint here will help guest supply service faster.
Quoted 5 lines
Quoted 1 line
I see.
I'll update the patch with your suggestion later.Hi, Zhang
Here is the idea and pseudo code to handle the "drop packet".```
ret =packet_enqueue
(1) ret == 0
compare connection
(2) ret == -1
send packet
(3) ret == queue insertion fail
do checkpoint
send all queued primary packets
remove all queued secondary packets ```
Quoted 1 line
Do you have any comment for the handling?
Looks good for me.
Thanks
Zhang Chen
Quoted 94 lines
Thanks
DerekQuoted 7 lines
Quoted 7 lines
Quoted 5 lines
It seems that frequent checkpoints caused by the full queue (busy network) instead of different
response packets may harm the high speed network (10 Gbps or higher) performance dramatically.Yes, maybe I can send a patch to make user adjust queue size depend on
it's own environment.
Quoted 82 lines
Quoted 2 lines
But with larger queue size, colo-compare will spend much time to do compare packet when network Is real busy status.
Thank you. The user-configurable queue size will be very helpful.
Thanks.
Derek SuQuoted 45 lines
Thanks
Zhang ChenQuoted 41 lines
Thanks
DerekQuoted 14 lines
Quoted 14 lines
Signed-off-by Derek Su <[email protected]>--- net/colo-compare.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-)net/colo-compare.c index renamed from net/colo-compare.c +0
-0
7ee17f2cf8..cdd87b2aa8100644net/colo-compare.c +0
-0@@ -120,6 +120,10 @@ enum { SECONDARY_IN, };Quoted 8 lines
Quoted 8 lines
net/colo-compare.c +4
-0+static const char *colo_mode[] = { + [PRIMARY_IN ] ="primary" , + [SECONDARY_IN ] ="secondary" , }; static int compare_chr_send(CompareState *s, const uint8_t *buf, @@ -215,6 +219,7 @@ static int packet_enqueue(CompareState *s, int mode,Connection
**con)
Quoted 9 lines
Quoted 9 lines
ConnectionKey key; Packet *pkt = NULL; Connection *conn;+ int ret;
if (mode == PRIMARY_IN) { pkt = packet_new(s->pri_rs.buf, @@ -243,16 +248,18 @@static int packet_enqueue(CompareState *s, int mode, Connection
**con)
Quoted 4 lines
Quoted 4 lines
} if (mode == PRIMARY_IN) {- if (!colo_insert_packet(&conn->primary_list, pkt, &conn-
pack)) {
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
net/colo-compare.c static int packet_enqueue(CompareState *s, int mode, Connection **con)
Quoted 7 lines
net/colo-compare.c +2
-4-error_report ("colo compare primary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->primary_list, pkt, + &conn->pack); }else { - if (!colo_insert_packet(&conn->secondary_list, pkt, &conn-sack)) {
Quoted 20 lines
Quoted 20 lines
Quoted 20 lines
Quoted 19 lines
net/colo-compare.c +10
-4-error_report ("colo compare secondary queue size too big," -"drop packet" ); - } + ret =colo_insert_packet (&conn->secondary_list, pkt, + &conn->sack); } + +if (!ret) { +error_report ("colo compare %s queue size too big," +"drop packet" , colo_mode[mode]); +packet_destroy (pkt,NULL ); + pkt =NULL ; + } + *con = conn;return 0 ; --2.17.1