Instead make ehci_execute and ehci_fill_queue return the again value.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Since with the ehci and xhci controllers a single packet can be larger
then maxpacketsize, it is possible for the result of a single packet
to be both having transferred some data as well as the transfer to have
an error.
An example would be an input transfer from a bulk endpoint successfully
receiving 1 or more maxpacketsize packets from the device, followed
by a packet signalling halt.
While already touching all the devices and controllers handle_packet /
handle_data / handle_control code, also change the return type of
these functions to void, solely storing the status in the packet. To
make the code paths for regular versus async packet handling more
uniform.
This patch unfortunately is somewhat invasive, since makeing the qemu
usb core deal with this requires changes everywhere. This patch only
prepares the usb core for this, all the hcd / device changes are done
in such a way that there are no functional changes.
This patch has been tested with uhci and ehci hcds, together with usb-audio,
usb-hid and usb-storage devices, as well as with usb-redir redirection
with a wide variety of real devices.
Note that there is usually no need to directly set packet->actual_length
form devices handle_data callback, as that is done by usb_packet_copy()
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Guard against re-definition of EHCI_DEBUG. Allows for turning on of debug info
from configure (using --qemu-extra-cflags="-DEHCI_DEBUG=1") rather than source
code hacking.
Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Seperate the PCI stuff from the EHCI components. Extracted the PCIDevice
out into a new wrapper struct to make EHCIState non-PCI-specific. Seperated
tho non PCI init component out into a seperate "common" init function.
Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Pull the DMAContext for the PCI DMA out at device init time and put it into
the device state. Use dma_memory_read/write() instead of pci specific versions.
Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The capabilities register and operational register offsets can vary from one
EHCI implementation to the next. Parameterise accordingly.
Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
ehci was already testing for this, and we depend in various places
on no devices doing this, so lets move the check for this to the
usb core.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
After a short-not-ok packet ending short, we should not advance the queue.
Move enforcing this to the core, rather then handling it in the hcd code.
This may result in the queue now actually containing multiple input packets
(which would not happen before), and this requires special handling in
combination with pipelining, so disable pipleining for input endpoints
(for now).
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
hcds which queue up more then one packet at once (uhci, ehci and xhci),
must clear the queue after an error which has caused the queue to halt.
Currently this is handled as a special case inside the hcd code, this
patch instead adds an USB_RET_REMOVE_FROM_QUEUE packet result code, teaches
the 3 hcds about this and moves the clearing of the queue on a halt into
the USB core.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This can be used by usb-device code which wishes to process an entire endpoint
queue at once, to do this the usb-device code returns USB_RET_ADD_TO_QUEUE
from its handle_data class method and defines a flush_ep_queue class method
to call when the hcd is done queuing up packets.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
If the guest is using multiple transfers to try and keep the usb bus busy /
used at maximum efficiency, currently we would see / do the following:
1) submit transfer 1 to the device
2) submit transfer 2 to the device
3) report transfer 1 completion to guest
4) report transfer 2 completion to guest
5) submit transfer 1 to the device
6) report transfer 1 completion to guest
7) submit transfer 2 to the device
8) report transfer 2 completion to guest
etc.
So after the initial submission we would effectively only have 1 transfer
in flight, rather then 2. This is caused by us not checking the queue for
addition of new transfers by the guest (ie the resubmission of a recently
finished transfer), while waiting for a pending transfer to complete.
This patch does add a check for this, changing the sequence to:
1) submit transfer 1 to the device
2) submit transfer 2 to the device
3) report transfer 1 completion to guest
4) submit transfer 1 to the device
5) report transfer 2 completion to guest
6) submit transfer 2 to the device
etc.
Thus keeping 2 transfers in flight (most of the time, and always 1),
as intended by the guest.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
For ctrl endpoints Windows (atleast Win7) creates circular td lists, so far
these were not a problem because we would stop filling the queue if altnext
was set. Since further patches in this patchset remove the altnext check this
does become a problem and we need detection for going in circles.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Often the guest will queue up new packets in response to a packet, in the
async schedule with its IOC flag set, completing. By speeding up the
frame-timer, we notice these new packets earlier. This increases the
speed (MB/s) of a Linux guest reading from a USB mass storage device by a
factor of 1.15 on top of the "Improve latency of interrupt delivery"
speed-ups, both with and without input pipelining enabled.
I've not tested the speed-up of this patch without the
"Improve latency of interrupt delivery" patch.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
While doing various performance tests of reading from USB mass storage devices
I noticed the following::
1) When an async handled packet completes, we don't immediately report an
interrupt to the guest, instead we wait for the frame-timer to run and
report it from there
2) If 1) has been fixed and an async handled packet takes a while to complete,
then async_stepdown will become a high value, which means that there
will be a large latency before any new packets queued by the guest in
response to the interrupt get seen
1) was done deliberately as part of commit f0ad01f92:
http://www.kraxel.org/cgit/qemu/commit/?h=usb.57&id=f0ad01f92ca02eee7cadbfd225c5de753ebd5fce
Since setting the interrupt immediately on async packet completion was causing
issues with Linux guests, I believe this recently fixed Linux bug explains
why this is happening:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=361aabf395e4a23cf554cf4ec0c0c6963b8beb01
Note that we can *not* count on this fix being present in all Linux guests!
I was hoping that the recently added support for Interrupt Threshold Control
would fix the issues with Linux guests, but adding a simple ehci_commit_irq()
call to ehci_async_bh() still caused problems with Linux guests.
The problem is, that when doing ehci_commit_irq() from ehci_async_bh(),
the "old" frindex value is used to calculate usbsts_frindex, and when
the frame-timer then runs possibly very shortly after ehci_async_bh(),
it increases the frame-timer, and thus any interrupts raised from that
frame-timer run, will also get reported to the guest immediately, rather
then being delayed to the next frame-timer run.
Luckily the solution for this is simple, this means that we need to
increase frindex before calling ehci_commit_irq() from ehci_async_bh(),
which in the end boils down to simple calling ehci_frame_timer() instead
of ehci_async_bh() from the bh.
This may seem like it causes a lot of extra work to be done, but this
is not true. Any work done from the frame-timer processing the periodic
schedule is work which then does not need to be done the next time the
frame timer runs, also the frame-timer will re-arm itself at (possibly)
a later time then it was armed for saving a vmexit at that time.
As an additional advantage moving to simply calling the frame-timer also
fixes 2) as the packet completion will set async_stepdown to 0, and the
re-arming of the timer with an async_stepdown of 0 ensures that any
newly queued up packets get seen in a reasonable amount of time.
This improves the speed (MB/s) of a Linux guest reading from a USB mass
storage device by a factor of 1.5 - 1.7 with input pipelining disabled,
and by a factor of 1.8 with input pipelining enabled.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
According to 4.15.1.2 an interrupt must be raised when a short packet
is received. If we don't do this it may take a significant time for
the guest to notice a short trasnfer has completed, since only the last td
will have its IOC flag set, and a short transfer may complete in an earlier
packet.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This field is used in some places to track the tbytes field of the token, but
in other places the field is used directly, use it directly everywhere for
consistency.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
target_phys_addr_t is unwieldly, violates the C standard (_t suffixes are
reserved) and its purpose doesn't match the name (most target_phys_addr_t
addresses are not target specific). Replace it with a finger-friendly,
standards conformant hwaddr.
Outstanding patchsets can be fixed up with the command
git rebase -i --exec 'find -name "*.[ch]"
| xargs s/target_phys_addr_t/hwaddr/g' origin
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
There are several issues with our handling of the MULT epcap field
of interrupt qhs, which this patch fixes.
1) When we don't execute a transaction because of the transaction counter
being 0, p->async stays EHCI_ASYNC_NONE, and the next time we process the
same qtd we hit an assert in ehci_state_fetchqtd because of this. Even though
I believe that this is caused by 3 below, this patch still removes the assert,
as that can still happen without 3, when multiple packets are queued for the
same interrupt ep.
2) We only *check* the transaction counter from ehci_state_execute, any
packets queued up by fill_queue bypass this check. This is fixed by not calling
fill_queue for interrupt packets.
3) Some versions of Windows set the MULT field of the qh to 0, which is a
clear violation of the EHCI spec, but still they do it. This means that we
will never execute a qtd for these, making interrupt ep-s on USB-2 devices
not work, and after recent changes, triggering 1).
So far we've stored the transaction counter in our copy of the mult field,
but with this beginnig at 0 already when dealing with these version of windows
this won't work. So this patch adds a transact_ctr field to our qh struct,
and sets this to the MULT field value on fetchqh. When the MULT field value
is 0, we set it to 4. Assuming that windows gets way with setting it to 0,
by the actual hardware going horizontal on a 1 -> 0 transition, which will
give it 4 transactions (MULT goes from 0 - 3).
Note that we cannot stop on detecting the 1 -> 0 transition, as our decrement
of the transaction counter, and checking for it are done in 2 different places.
Reported-by: Shawn Starr <shawn.starr@rogers.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
When removing unseen queue-heads from the async queue list, we should not
set the seen flag to 0, as this may cause them to be removed by
ehci_queues_rip_unused() during the next call to ehci_advance_async_state()
if the timer is late or running at a low frequency.
Note:
1) This *may* have caused the instant unlink / relinks described in commit
9bc3a3a216
2) Rather then putting more if-s inside ehci_queues_rip_unused, this patch
instead introduces a new ehci_queues_rip_unseen function.
3) This patch also makes it save to call ehci_queues_rip_unseen() multiple
times, which gets used in the folluw up patch titled:
"ehci: Walk async schedule before and after migration"
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The Linux ehci isoc scheduling code fills the entire schedule ahead of
time minus 80 frames. If we make a large jump in where we are in the
schedule, ie 40 frames, then the scheduler all of a sudden will only have
40 frames left to work in, causing it to fail packet submissions
with error -27 (-EFBIG).
Changes in v2:
-Don't hardcode a maximum number of frames to process in one tick, instead:
-Process a minimum number of frames to ensure we do eventually catch up
-Stop (after the minimum number) when the guest has requested an irq
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
If Interrupt Threshold Control is 8 or a multiple of 8, then
s->usbsts_frindex can become exactly 0x4000, at which point
(s->usbsts_frindex > s->frindex) will never become true, as
s->usbsts_frindex will not be lowered / reset in this case.
This patch fixes this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Also register different memory regions for capabilities,
operational registers and port status registers. Create
separate tracepoints for operational regs and port status
regs. Ditch a bunch of sanity checks because the memory
core will do this for us now.
Offloading the byte, word and dword access handling to the
memory core also has the side effect of fixing ehci register
access on bigendian hosts.
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Since my previous comment said "Should never happen", I tried changing the
next line to an assert(0), which did not go well, which as the new comments
explains is logical if you think about it for a moment.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
USB_RET_PROCERR can be triggered by the guest (by for example requesting more
then BUFFSIZE bytes), so don't assert on it.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Currently each time we try to execute a NAK-ed packet we redo
ehci_init_transfer, and usb_packet_map, re-allocing (without freeing) the
sg list every time.
This patch fixes this, it does this by introducing another async state, so
that we also properly cleanup a NAK-ed packet on cancel.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
make qemu_queue_{cancel,reset} return the number of packets released,
so the caller can figure whenever there have been active packets even
though there shouldn't have been any. Add tracepoint to log this.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reported packets which have completed before being cancelled as such to the
host. Note that the new code path this patch adds is untested since it I've
been unable to actually trigger the race which needs this code path.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
-combine the qh check with the check for devaddr changes
-also ensure that p gets set to NULL when the queue gets cancelled on
devaddr change, which was not done properly before this patch
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
This reverts commit 9bc3a3a216, which got
added to fix an issue where the real, underlying cause was not stopping
the ep queue on an error.
Now that the underlying cause is fixed by the "usb: Halt ep queue and
cancel pending packets on a packet error" patch, the "don't flush" fix
is no longer needed.
Not only is it not needed, it causes us to see cancellations (unlinks)
done by the Linux EHCI driver too late, which in combination with the new
usb-core packet-id generation where qtd addresses are used as ids, causes
duplicate ids for in flight packets.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
If an (emulated) usb-device tries to write more data to a packet then
its iov len, this will trigger an assert in usb_packet_copy(), and if
a driver somehow circumvents that check and writes more data to the
iov then there is space, we have a much bigger problem then not correctly
reporting babble to the guest.
In practice babble will only happen with (real) redirected devices, and there
both the usb-host os and the qemu usb-device code already check for it.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
One of the recent changes (likely the addition of queuing support) has broken
interrupt endpoints, this patch fixes this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
ehci_state_executing does not need to check for p->usb_status == USB_RET_ASYNC
or USB_RET_PROCERR, since ehci_execute_complete already does a similar check
and will trigger an assert if either value is encountered.
USB_RET_ASYNC should never be the packet status when execute_complete runs
for obvious reasons, and USB_RET_PROCERR is only used by ehci_state_execute /
ehci_execute not by ehci_state_executing / ehci_execute_complete.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
ehci_qh_do_overlay() already calls ehci_flush_qh() before it returns, calling
it twice is useless.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
After the "ehci: Print a warning when a queue unexpectedly contains packets
on cancel" commit. Under certain reproducable conditions I was getting the
following message: "EHCI: Warning queue not empty on queue reset".
After aprox. 8 hours of debugging I've finally found the cause. The Linux EHCI
driver has an IAAD watchdog, to work around certain EHCI hardware sometimes
not acknowledging the doorbell at all. This watchdog has a timeout of 10 ms,
which is less then the time between 2 runs through the async schedule when
async_stepdown is at its highest value.
Thus the watchdog can trigger, after which Linux clears the IAAD bit and
re-uses the QH. IOW we were not properly detecting the unlink of the qh, due
to us missing (ignoring for more then 10 ms) the IAAD command, which triggered
the warning.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
This patch adds IDs to usb packets. Those IDs are (a) supposed to be
unique for the lifecycle of a packet (from packet setup until the packet
is either completed or canceled) and (b) stable across migration.
uhci, ohci, ehci and xhci use the guest physical address of the transfer
descriptor for this.
musb needs a different approach because there is no transfer descriptor.
But musb also doesn't support pipelining, so we have never more than one
packet per endpoint in flight. So we go create an ID based on endpoint
and device address.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
For controllers which queue up more then 1 packet at a time, we must halt the
ep queue, and inside the controller code cancel all pending packets on an
error.
There are multiple reasons for this:
1) Guests expect the controllers to halt ep queues on error, so that they
get the opportunity to cancel transfers which the scheduled after the failing
one, before processing continues
2) Not cancelling queued up packets after a failed transfer also messes up
the controller state machine, in the case of EHCI causing the following
assert to trigger: "assert(p->qtdaddr == q->qtdaddr)" at hcd-ehci.c:2075
3) For bulk endpoints with pipelining enabled (redirection to a real USB
device), we must cancel all the transfers after this a failed one so that:
a) If they've completed already, they are not processed further causing more
stalls to be reported, originating from the same failed transfer
b) If still in flight, they are cancelled before the guest does
a clear stall, otherwise the guest and device can loose sync!
Note this patch only touches the ehci and uhci controller changes, since AFAIK
no other controllers actually queue up multiple transfer. If I'm wrong on this
other controllers need to be updated too!
Also note that this patch was heavily tested with the ehci code, where I had
a reproducer for a device causing a transfer to fail. The uhci code is not
tested with actually failing transfers and could do with a thorough review!
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This fixes linux guests started without any USB devices not seeing newly
plugged devices until "lsusb" is done inside the guest.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
First, not all interrupts are subject to Interrupt Threshold Control,
some of them must be delivered without delay.
Second, Interrupt Threshold Control state must be part of vmstate,
otherwise we might loose IRQs on migration.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
