* Re: [Qemu-devel] [Patch 1/1] net/net: Allocating Large sized arrays to heap
2016-03-12 8:39 [Qemu-devel] [Patch 1/1] net/net: Allocating Large sized arrays to heap Pooja Dhannawat
@ 2016-03-14 1:44 ` Li Zhijian
2016-03-14 14:11 ` Eric Blake
2016-03-14 18:26 ` [Qemu-devel] [PATCH v2] net: " Pooja Dhannawat
2 siblings, 0 replies; 5+ messages in thread
From: Li Zhijian @ 2016-03-14 1:44 UTC (permalink / raw)
To: Pooja Dhannawat, qemu-devel
could you explain why we need this change ?
On 03/12/2016 04:39 PM, Pooja Dhannawat wrote:
> Signed-off-by: Pooja Dhannawat <dhannawatpooja1@gmail.com>
> ---
> net/net.c | 8 ++++++--
> 1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/net/net.c b/net/net.c
> index b0c832e..5399758 100644
> --- a/net/net.c
> +++ b/net/net.c
> @@ -709,16 +709,18 @@ ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
> static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
> int iovcnt, unsigned flags)
> {
> - uint8_t buf[NET_BUFSIZE];
> + uint8_t *buf;
> uint8_t *buffer;
> size_t offset;
>
> + buf = g_new(uint8_t, 1);
> +
> if (iovcnt == 1) {
> buffer = iov[0].iov_base;
> offset = iov[0].iov_len;
> } else {
> buffer = buf;
> - offset = iov_to_buf(iov, iovcnt, 0, buf, sizeof(buf));
> + offset = iov_to_buf(iov, iovcnt, 0, (uint8_t *)buf, sizeof(uint8_t));
> }
>
> if (flags & QEMU_NET_PACKET_FLAG_RAW && nc->info->receive_raw) {
> @@ -726,6 +728,8 @@ static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
> } else {
forget to release 'buf'
> return nc->info->receive(nc, buffer, offset);
> }
> +
> + g_free(buf);
> }
>
> ssize_t qemu_deliver_packet_iov(NetClientState *sender,
>
--
Best regards.
Li Zhijian
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [Qemu-devel] [Patch 1/1] net/net: Allocating Large sized arrays to heap
2016-03-12 8:39 [Qemu-devel] [Patch 1/1] net/net: Allocating Large sized arrays to heap Pooja Dhannawat
2016-03-14 1:44 ` Li Zhijian
@ 2016-03-14 14:11 ` Eric Blake
2016-03-14 14:49 ` Pooja Dhannawat
2016-03-14 18:26 ` [Qemu-devel] [PATCH v2] net: " Pooja Dhannawat
2 siblings, 1 reply; 5+ messages in thread
From: Eric Blake @ 2016-03-14 14:11 UTC (permalink / raw)
To: Pooja Dhannawat, qemu-devel
[-- Attachment #1: Type: text/plain, Size: 1761 bytes --]
On 03/12/2016 01:39 AM, Pooja Dhannawat wrote:
> Signed-off-by: Pooja Dhannawat <dhannawatpooja1@gmail.com>
> ---
> net/net.c | 8 ++++++--
> 1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/net/net.c b/net/net.c
> index b0c832e..5399758 100644
> --- a/net/net.c
> +++ b/net/net.c
> @@ -709,16 +709,18 @@ ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
> static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
> int iovcnt, unsigned flags)
> {
> - uint8_t buf[NET_BUFSIZE];
> + uint8_t *buf;
> uint8_t *buffer;
> size_t offset;
>
> + buf = g_new(uint8_t, 1);
If you're only going to malloc() one byte, it's more efficient to just
stack-allocate it:
uint8_t buf[1];
Did you mean to do:
buf = g_new(uint8_t, NET_BUFSIZE)
instead?
> +
> if (iovcnt == 1) {
> buffer = iov[0].iov_base;
> offset = iov[0].iov_len;
> } else {
> buffer = buf;
> - offset = iov_to_buf(iov, iovcnt, 0, buf, sizeof(buf));
> + offset = iov_to_buf(iov, iovcnt, 0, (uint8_t *)buf, sizeof(uint8_t));
This is wrong. The old code used NET_BUFSIZE bytes for iov_to_buf(),
the new code uses only one byte. By the way, sizeof(char) and
sizeof(uint8_t) are pointless in code, as they are strictly equivalent to 1.
I agree that the idea behind the patch is useful (NET_BUFSIZE is 68k,
which is too large for our goal of at most 4k stack allocation if we
never want to overflow a stack guard page), but you'll need a correct
working version of the patch.
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [Qemu-devel] [Patch 1/1] net/net: Allocating Large sized arrays to heap
2016-03-14 14:11 ` Eric Blake
@ 2016-03-14 14:49 ` Pooja Dhannawat
0 siblings, 0 replies; 5+ messages in thread
From: Pooja Dhannawat @ 2016-03-14 14:49 UTC (permalink / raw)
To: Eric Blake; +Cc: qemu-devel
[-- Attachment #1: Type: text/plain, Size: 1984 bytes --]
On Mon, Mar 14, 2016 at 7:41 PM, Eric Blake <eblake@redhat.com> wrote:
> On 03/12/2016 01:39 AM, Pooja Dhannawat wrote:
> > Signed-off-by: Pooja Dhannawat <dhannawatpooja1@gmail.com>
> > ---
> > net/net.c | 8 ++++++--
> > 1 file changed, 6 insertions(+), 2 deletions(-)
> >
> > diff --git a/net/net.c b/net/net.c
> > index b0c832e..5399758 100644
> > --- a/net/net.c
> > +++ b/net/net.c
> > @@ -709,16 +709,18 @@ ssize_t qemu_send_packet_raw(NetClientState *nc,
> const uint8_t *buf, int size)
> > static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec
> *iov,
> > int iovcnt, unsigned flags)
> > {
> > - uint8_t buf[NET_BUFSIZE];
> > + uint8_t *buf;
> > uint8_t *buffer;
> > size_t offset;
> >
> > + buf = g_new(uint8_t, 1);
>
> If you're only going to malloc() one byte, it's more efficient to just
> stack-allocate it:
>
> uint8_t buf[1];
>
> Did you mean to do:
>
> buf = g_new(uint8_t, NET_BUFSIZE)
>
> instead?
>
> Yes, my bad, I meant NET_BUFSIZE.
> > +
> > if (iovcnt == 1) {
> > buffer = iov[0].iov_base;
> > offset = iov[0].iov_len;
> > } else {
> > buffer = buf;
> > - offset = iov_to_buf(iov, iovcnt, 0, buf, sizeof(buf));
> > + offset = iov_to_buf(iov, iovcnt, 0, (uint8_t *)buf,
> sizeof(uint8_t));
>
> This is wrong. The old code used NET_BUFSIZE bytes for iov_to_buf(),
> the new code uses only one byte. By the way, sizeof(char) and
> sizeof(uint8_t) are pointless in code, as they are strictly equivalent to
> 1.
>
> I agree that the idea behind the patch is useful (NET_BUFSIZE is 68k,
> which is too large for our goal of at most 4k stack allocation if we
> never want to overflow a stack guard page), but you'll need a correct
> working version of the patch.
>
> Yes, I will correct it and send next version of patch.
> --
> Eric Blake eblake redhat com +1-919-301-3266
> Libvirt virtualization library http://libvirt.org
>
>
[-- Attachment #2: Type: text/html, Size: 3163 bytes --]
^ permalink raw reply [flat|nested] 5+ messages in thread
* [Qemu-devel] [PATCH v2] net: Allocating Large sized arrays to heap
2016-03-12 8:39 [Qemu-devel] [Patch 1/1] net/net: Allocating Large sized arrays to heap Pooja Dhannawat
2016-03-14 1:44 ` Li Zhijian
2016-03-14 14:11 ` Eric Blake
@ 2016-03-14 18:26 ` Pooja Dhannawat
2 siblings, 0 replies; 5+ messages in thread
From: Pooja Dhannawat @ 2016-03-14 18:26 UTC (permalink / raw)
To: qemu-devel
nc_sendv_compat has a huge stack usage of 69680 bytes approx.
Moving large arrays to heap to reduce stack usage.
Signed-off-by: Pooja Dhannawat <dhannawatpooja1@gmail.com>
---
net/net.c | 14 ++++++++++----
1 file changed, 10 insertions(+), 4 deletions(-)
diff --git a/net/net.c b/net/net.c
index b0c832e..89589c0 100644
--- a/net/net.c
+++ b/net/net.c
@@ -709,23 +709,29 @@ ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
int iovcnt, unsigned flags)
{
- uint8_t buf[NET_BUFSIZE];
+ uint8_t *buf;
uint8_t *buffer;
size_t offset;
+ ssize_t ret;
+
+ buf = g_new(uint8_t, NET_BUFSIZE);
if (iovcnt == 1) {
buffer = iov[0].iov_base;
offset = iov[0].iov_len;
} else {
buffer = buf;
- offset = iov_to_buf(iov, iovcnt, 0, buf, sizeof(buf));
+ offset = iov_to_buf(iov, iovcnt, 0, (uint8_t *)buf, NET_BUFSIZE);
}
if (flags & QEMU_NET_PACKET_FLAG_RAW && nc->info->receive_raw) {
- return nc->info->receive_raw(nc, buffer, offset);
+ ret = nc->info->receive_raw(nc, buffer, offset);
} else {
- return nc->info->receive(nc, buffer, offset);
+ ret = nc->info->receive(nc, buffer, offset);
}
+
+ g_free(buf);
+ return ret;
}
ssize_t qemu_deliver_packet_iov(NetClientState *sender,
--
2.5.0
^ permalink raw reply related [flat|nested] 5+ messages in thread