From: "Michael S. Tsirkin" <mst@redhat.com>
To: qemu-devel@nongnu.org
Cc: Roman Kagan <rvkagan@yandex-team.ru>,
Peter Maydell <peter.maydell@linaro.org>
Subject: [PULL 01/10] pci/shpc: don't push attention button when ejecting powered-off device
Date: Sun, 17 Jan 2021 06:46:18 -0500 [thread overview]
Message-ID: <20210117114519.539647-2-mst@redhat.com> (raw)
In-Reply-To: <20210117114519.539647-1-mst@redhat.com>
From: Roman Kagan <rvkagan@yandex-team.ru>
When the slot is in steady powered-off state and the device is being
removed, there's no need to press the attention button. Nor is it
mandated by the Standard Hot-Plug Controller Specification, Rev. 1.0.
Moreover it confuses the guest, Linux in particular, as it assumes that
the attention button pressed in this state indicates that the device has
been inserted and will need to be powered on. Therefore it transitions
the slot into BLINKING_ON state for 5 seconds, and discovers at the end
that no device is actually inserted:
... unplug request
[12685.451329] shpchp 0000:01:00.0: Button pressed on Slot(2)
[12685.455478] shpchp 0000:01:00.0: PCI slot #2 - powering off due to button press
... in 5 seconds OS powers off the slot, QEMU ejects the device
[12690.632282] shpchp 0000:01:00.0: Latch open on Slot(2)
... excessive button press in steady powered-off state
[12690.634267] shpchp 0000:01:00.0: Button pressed on Slot(2)
[12690.636256] shpchp 0000:01:00.0: Card not present on Slot(2)
... the last button press spawns powering on the slot
[12690.638909] shpchp 0000:01:00.0: PCI slot #2 - powering on due to button press
... in 5 more seconds attempt to power on discovers empty slot
[12695.735986] shpchp 0000:01:00.0: No adapter on slot(2)
Worse, if the real device insertion happens within 5 seconds from the
apparent completion of the previous device removal (signaled via
DEVICE_DELETED event), the new button press will be interpreted as the
cancellation of that misguided powering on:
[13448.965295] shpchp 0000:01:00.0: Button pressed on Slot(2)
[13448.969430] shpchp 0000:01:00.0: PCI slot #2 - powering off due to button press
[13454.025107] shpchp 0000:01:00.0: Latch open on Slot(2)
[13454.027101] shpchp 0000:01:00.0: Button pressed on Slot(2)
[13454.029165] shpchp 0000:01:00.0: Card not present on Slot(2)
... the excessive button press spawns powering on the slot
... device has already been ejected by QEMU
[13454.031949] shpchp 0000:01:00.0: PCI slot #2 - powering on due to button press
... new device is inserted in the slot
[13456.861545] shpchp 0000:01:00.0: Latch close on Slot(2)
... valid button press arrives before 5 s since the wrong one
[13456.864894] shpchp 0000:01:00.0: Button pressed on Slot(2)
[13456.869211] shpchp 0000:01:00.0: Card present on Slot(2)
... the valid button press is counted as cancellation of the wrong one
[13456.873173] shpchp 0000:01:00.0: Button cancel on Slot(2)
[13456.877101] shpchp 0000:01:00.0: PCI slot #2 - action canceled due to button press
As a result, the newly inserted device isn't brought up by the guest.
Avoid this situation by not pushing the attention button when the device
in the slot is in powered-off state and is being ejected.
FWIW pcie implementation doesn't suffer from this problem.
Signed-off-by: Roman Kagan <rvkagan@yandex-team.ru>
Message-Id: <20201102053750.2281818-1-rvkagan@yandex-team.ru>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
hw/pci/shpc.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/hw/pci/shpc.c b/hw/pci/shpc.c
index 4786a44996..28e62174c4 100644
--- a/hw/pci/shpc.c
+++ b/hw/pci/shpc.c
@@ -300,7 +300,6 @@ static void shpc_slot_command(SHPCDevice *shpc, uint8_t target,
shpc_set_status(shpc, slot, SHPC_SLOT_STATUS_PRSNT_EMPTY,
SHPC_SLOT_STATUS_PRSNT_MASK);
shpc->config[SHPC_SLOT_EVENT_LATCH(slot)] |=
- SHPC_SLOT_EVENT_BUTTON |
SHPC_SLOT_EVENT_MRL |
SHPC_SLOT_EVENT_PRESENCE;
}
@@ -566,7 +565,6 @@ void shpc_device_unplug_request_cb(HotplugHandler *hotplug_dev,
return;
}
- shpc->config[SHPC_SLOT_EVENT_LATCH(slot)] |= SHPC_SLOT_EVENT_BUTTON;
state = shpc_get_status(shpc, slot, SHPC_SLOT_STATE_MASK);
led = shpc_get_status(shpc, slot, SHPC_SLOT_PWR_LED_MASK);
if (state == SHPC_STATE_DISABLED && led == SHPC_LED_OFF) {
@@ -577,6 +575,8 @@ void shpc_device_unplug_request_cb(HotplugHandler *hotplug_dev,
shpc->config[SHPC_SLOT_EVENT_LATCH(slot)] |=
SHPC_SLOT_EVENT_MRL |
SHPC_SLOT_EVENT_PRESENCE;
+ } else {
+ shpc->config[SHPC_SLOT_EVENT_LATCH(slot)] |= SHPC_SLOT_EVENT_BUTTON;
}
shpc_set_status(shpc, slot, 0, SHPC_SLOT_STATUS_66);
shpc_interrupt_update(pci_hotplug_dev);
--
MST
next prev parent reply other threads:[~2021-01-17 11:50 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-17 11:46 [PULL 00/10] pc,pci,virtio: fixes, features Michael S. Tsirkin
2021-01-17 11:46 ` Michael S. Tsirkin [this message]
2021-01-17 11:46 ` [PULL 02/10] vhost-user-fs: add the "bootindex" property Michael S. Tsirkin
2021-01-17 11:46 ` [PULL 03/10] acpi: Allow DSDT acpi table changes Michael S. Tsirkin
2021-01-17 11:46 ` [PULL 04/10] acpi: Fix unmatched expected DSDT.pxb file Michael S. Tsirkin
2021-01-17 11:46 ` [PULL 05/10] acpi: Add addr offset in build_crs Michael S. Tsirkin
2021-01-17 11:46 ` [PULL 06/10] acpi/gpex: Inform os to keep firmware resource map Michael S. Tsirkin
2021-01-17 11:46 ` [PULL 07/10] acpi/gpex: Exclude pxb's resources from PCI0 Michael S. Tsirkin
2021-01-17 11:46 ` [PULL 08/10] Kconfig: Compile PXB for ARM_VIRT Michael S. Tsirkin
2021-01-17 11:46 ` [PULL 09/10] acpi: Enable pxb unit-test for ARM virt machine Michael S. Tsirkin
2021-01-17 11:46 ` [PULL 10/10] acpi: Update _DSM method in expected files Michael S. Tsirkin
2021-01-18 10:28 ` [PULL 00/10] pc,pci,virtio: fixes, features Peter Maydell
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=20210117114519.539647-2-mst@redhat.com \
--to=mst@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=qemu-devel@nongnu.org \
--cc=rvkagan@yandex-team.ru \
/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).