From: Pavel Dovgalyuk <pavel.dovgalyuk@ispras.ru>
To: Jason Wang <jasowang@redhat.com>
Cc: qemu-devel <qemu-devel@nongnu.org>,
"Dr . David Alan Gilbert" <dgilbert@redhat.com>
Subject: Re: [PATCH] hw/net: store timers for e1000 in vmstate
Date: Wed, 27 Oct 2021 10:07:02 +0300 [thread overview]
Message-ID: <7efd0945-c885-6bda-96cd-47ce56eae85d@ispras.ru> (raw)
In-Reply-To: <CACGkMEuwW6miZ0GAqFC7WFK+1oOAYoz7x-zC-+Z-u0A_n3x0aA@mail.gmail.com>
On 27.10.2021 07:05, Jason Wang wrote:
> On Tue, Oct 26, 2021 at 6:36 PM Pavel Dovgalyuk
> <pavel.dovgalyuk@ispras.ru> wrote:
>>
>> Setting timers randomly when vmstate is loaded breaks
>> execution determinism.
>> Therefore this patch allows saving mit and autoneg timers
>> for e1000. It makes execution deterministic and allows
>> snapshotting and reverse debugging in icount mode.
>>
>> Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
>> ---
>> hw/net/e1000.c | 61 ++++++++++++++++++++++++++++++++++++++++++++++----------
>> 1 file changed, 50 insertions(+), 11 deletions(-)
>>
>> diff --git a/hw/net/e1000.c b/hw/net/e1000.c
>> index a30546c5d5..2f706f7298 100644
>> --- a/hw/net/e1000.c
>> +++ b/hw/net/e1000.c
>> @@ -37,6 +37,7 @@
>> #include "qemu/iov.h"
>> #include "qemu/module.h"
>> #include "qemu/range.h"
>> +#include "sysemu/replay.h"
>>
>> #include "e1000x_common.h"
>> #include "trace.h"
>> @@ -1407,7 +1408,7 @@ static int e1000_pre_save(void *opaque)
>> * complete auto-negotiation immediately. This allows us to look
>> * at MII_SR_AUTONEG_COMPLETE to infer link status on load.
>> */
>> - if (nc->link_down && have_autoneg(s)) {
>> + if (replay_mode == REPLAY_MODE_NONE && nc->link_down && have_autoneg(s)) {
>> s->phy_reg[PHY_STATUS] |= MII_SR_AUTONEG_COMPLETE;
>> }
>>
>> @@ -1438,22 +1439,12 @@ static int e1000_post_load(void *opaque, int version_id)
>> s->mac_reg[TADV] = 0;
>> s->mit_irq_level = false;
>> }
>> - s->mit_ide = 0;
>> - s->mit_timer_on = true;
>> - timer_mod(s->mit_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + 1);
>>
>> /* nc.link_down can't be migrated, so infer link_down according
>> * to link status bit in mac_reg[STATUS].
>> * Alternatively, restart link negotiation if it was in progress. */
>> nc->link_down = (s->mac_reg[STATUS] & E1000_STATUS_LU) == 0;
>>
>> - if (have_autoneg(s) &&
>> - !(s->phy_reg[PHY_STATUS] & MII_SR_AUTONEG_COMPLETE)) {
>> - nc->link_down = false;
>> - timer_mod(s->autoneg_timer,
>> - qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + 500);
>> - }
>
> So we won't get those timers armed after migration unconditionally. Is
> this intended?
Not really. I think there could be several solutions:
1. Save some flag to distinguish between old and new state.
2. Use deterministic version for icount (or even record/replay) mode only.
3. Check machine type to change the behavior (as Dave proposed)
>
> Thanks
>
>> -
>> s->tx.props = s->mig_props;
>> if (!s->received_tx_tso) {
>> /* We received only one set of offload data (tx.props)
>> @@ -1472,6 +1463,13 @@ static int e1000_tx_tso_post_load(void *opaque, int version_id)
>> return 0;
>> }
>>
>> +static int e1000_mit_timer_post_load(void *opaque, int version_id)
>> +{
>> + E1000State *s = opaque;
>> + s->mit_timer_on = true;
>> + return 0;
>> +}
>> +
>> static bool e1000_mit_state_needed(void *opaque)
>> {
>> E1000State *s = opaque;
>> @@ -1493,6 +1491,21 @@ static bool e1000_tso_state_needed(void *opaque)
>> return chkflag(TSO);
>> }
>>
>> +static bool e1000_mit_timer_needed(void *opaque)
>> +{
>> + E1000State *s = opaque;
>> +
>> + return s->mit_timer_on;
>> +}
>> +
>> +static bool e1000_autoneg_timer_needed(void *opaque)
>> +{
>> + E1000State *s = opaque;
>> +
>> + return have_autoneg(s)
>> + && !(s->phy_reg[PHY_STATUS] & MII_SR_AUTONEG_COMPLETE);
>> +}
>> +
>> static const VMStateDescription vmstate_e1000_mit_state = {
>> .name = "e1000/mit_state",
>> .version_id = 1,
>> @@ -1541,6 +1554,30 @@ static const VMStateDescription vmstate_e1000_tx_tso_state = {
>> }
>> };
>>
>> +static const VMStateDescription vmstate_e1000_mit_timer = {
>> + .name = "e1000/mit_timer",
>> + .version_id = 1,
>> + .minimum_version_id = 1,
>> + .needed = e1000_mit_timer_needed,
>> + .post_load = e1000_mit_timer_post_load,
>> + .fields = (VMStateField[]) {
>> + VMSTATE_TIMER_PTR(mit_timer, E1000State),
>> + VMSTATE_UINT32(mit_ide, E1000State),
>> + VMSTATE_END_OF_LIST()
>> + }
>> +};
>> +
>> +static const VMStateDescription vmstate_e1000_autoneg_timer = {
>> + .name = "e1000/autoneg_timer",
>> + .version_id = 1,
>> + .minimum_version_id = 1,
>> + .needed = e1000_autoneg_timer_needed,
>> + .fields = (VMStateField[]) {
>> + VMSTATE_TIMER_PTR(autoneg_timer, E1000State),
>> + VMSTATE_END_OF_LIST()
>> + }
>> +};
>> +
>> static const VMStateDescription vmstate_e1000 = {
>> .name = "e1000",
>> .version_id = 2,
>> @@ -1622,6 +1659,8 @@ static const VMStateDescription vmstate_e1000 = {
>> &vmstate_e1000_mit_state,
>> &vmstate_e1000_full_mac_state,
>> &vmstate_e1000_tx_tso_state,
>> + &vmstate_e1000_mit_timer,
>> + &vmstate_e1000_autoneg_timer,
>> NULL
>> }
>> };
>>
>
prev parent reply other threads:[~2021-10-27 7:08 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-10-26 10:31 [PATCH] hw/net: store timers for e1000 in vmstate Pavel Dovgalyuk
2021-10-26 17:02 ` Dr. David Alan Gilbert
2021-10-27 4:05 ` Jason Wang
2021-10-27 7:07 ` Pavel Dovgalyuk [this message]
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=7efd0945-c885-6bda-96cd-47ce56eae85d@ispras.ru \
--to=pavel.dovgalyuk@ispras.ru \
--cc=dgilbert@redhat.com \
--cc=jasowang@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).