CVS log for src/sys/dev/pci/vioscsi.c

Up to [cvs.NetBSD.org] / src / sys / dev / pci

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

Pull up following revision(s) (requested by rin in ticket #914):

	sys/dev/pci/virtio_pci.c: revision 1.50
	sys/dev/pci/virtio_pci.c: revision 1.51
	sys/dev/pci/virtio_pci.c: revision 1.52
	sys/dev/pci/virtio_pci.c: revision 1.53
	sys/dev/pci/virtio_pci.c: revision 1.54
	sys/dev/pci/vioscsi.c: revision 1.35
	sys/dev/pci/ld_virtio.c: revision 1.33
	sys/dev/pci/ld_virtio.c: revision 1.34
	sys/dev/virtio/viocon.c: revision 1.9
	sys/dev/pci/ld_virtio.c: revision 1.35
	sys/dev/pci/virtio.c: revision 1.80
	sys/dev/pci/virtio.c: revision 1.81
	sys/dev/pci/virtio.c: revision 1.82
	sys/dev/pci/virtio_pci.c: revision 1.45
	sys/dev/pci/virtio_pci.c: revision 1.46
	sys/dev/pci/virtio_pci.c: revision 1.47
	sys/dev/pci/virtio_pci.c: revision 1.48
	sys/dev/pci/virtio_pci.c: revision 1.49
	sys/dev/pci/virtio.c: revision 1.79

A single SCSI request may require multiple slots in the virtio queue
but the queue isn't sized for the theoretical maximum. So just use
XS_BUSY to pace the scsipi layer.

viocon(4): Show error in assertion failure.

virtio_read_device_config_le_[24]: Fix for {aarch64,arm}eb
Stop byte-swapping for big-endian aarch64 and arm to fix
corrupted read for, e.g., sc_taglen for vio9p(4).

As described as comments in virtio_pci.c, big-endian aarch64 and
armv7 (armeb) are somewhat special. At the moment, all supported
virtual/real machines are configured as little-endian, and only
CPU cores are switched to big-endian mode during early boot stage.

Most peripheral buses for armeb are LSB as a result, and their
default bus_space(9) functions swap byte-order. Therefore, PIOed
data from memory-mapped devices, as well as pci(4) ones, are
actually LSB, but observed as MSB for armeb.

Therefore, we should not swap byte-order further in
virtio_read_device_config_le_[24] on armeb.

See virtio_pci.c and sys/dev/fdt/virtio_mmio_fdt.c also.
fix spelling mistakes, mainly in comments and log messages.

Split KASSERT(A && B) into KASSERT(A); KASSERT(B).

Fix typo in error message.

Modify a confused expression in ld_virtio_attach().

VIRTIO_BLK_MIN_SEGMENTS should be the total number of non-data segments,
so I rename it to VIRTIO_BLK_CTRL_SEGMENTS.
PR kern/57981.

ld@virtio(4): Fix maximum size parameters.
- SEG_MAX is the maximum number of segments.
- SIZE_MAX is the maximum number of bytes in a single segment.
The maximum transfer size is, therefore, SEG_MAX * SIZE_MAX.
=> Don't add two extra segments in the dmamap vr_payload for the
   header and status -- we already have a separate dmamap vr_cmdsts
   for that.
=> Don't recalculate payload dmamap parameters based on division by
   NBPG, just use the ones specified by the host.
=> Allow SIZE_MAX below MAXPHYS as long as SIZE_MAX*SEG_MAX >=
   MAXPHYS.
Even though ldattach clamps ld->sc_maxxfer to MAXPHYS, make sure to
clamp it in ld_virtio_attach before ld_virtio_alloc_reqs since it
determines the dmamap sizes and bounce buffer allocation and there's
no sense in allocating those larger than ld will use anyway.
PR kern/58338

virtio_pci.c: Nix trailing whitespace.

virtio@pci: Unmap the BARs we actually mapped on >=1.0 detach.
PR kern/58335

virtio@pci: Nix useless sc_mapped_iosize.
This appears to have been a mistaken attempt to pass the subregion
length through to bus_space_unmap, which doesn't make any sense --
bus_space_subregion does not allocate anything and the resulting
handle must not be passed to bus_space_unmap.

Once we pass the size that was actually returned by pci_mapreg_map to
bus_space_unmap, which matches the size passed internally by
pci_mapreg_map to bus_space_map, sc_mapped_iosize is no longer used
and can be flushed.
Prompted by PR kern/58335.

virtio@pci: Don't detach on shutdown.
The virtio bus does not keep state in memory that needs to flushed to
persistent storage.  Typically DVF_DETACH_SHUTDOWN is only needed by,
e.g., disk drivers to flush disk caches (although really, vfs sync
should do that anyway, so maybe we should just nix the option
altogether).
Prompted by PR kern/58335.

virtio@pci: Zero sc_bars_iosize[i] on unmap.
No functional change intended, but this makes the code less brittle.
Prompted by PR kern/58335.

virtio_pci.c: Sort includes.
No functional change intended.

virtio_pci.c: Omit needless blank lines and commented code.
No functional change intended.

virtio_pci.c: Sprinkle KNF.
- No parameter names in function prototypes.
- Fix indentation.
- Break overlong lines.
- Blank line between declarations and rest of block.
virtio_pci.c: Sprinkle const.
No functional change intended.

virtio_pci.c: Use container_of, not cast.
virtio(4): Assert child attach finished before we allow queueing.
Should help catch mistakes like:
PR kern/58555: Kernel panic during boot when using viocon with PCIe

Pull up following revision(s) (requested by yamaguchi in ticket #139):

	sys/dev/pci/vioscsi.c: revision 1.31
	sys/dev/pci/vio9p.c: revision 1.10
	sys/dev/pci/vioscsi.c: revision 1.32
	sys/dev/pci/vio9p.c: revision 1.11
	sys/dev/pci/vioscsi.c: revision 1.33
	sys/dev/pci/ld_virtio.c: revision 1.31
	sys/dev/virtio/viocon.c: revision 1.6
	sys/dev/pci/vioscsi.c: revision 1.34
	sys/dev/pci/ld_virtio.c: revision 1.32
	sys/dev/virtio/viocon.c: revision 1.7
	sys/dev/virtio/viocon.c: revision 1.8
	sys/dev/pci/vioscsi.c: revision 1.36
	sys/dev/pci/virtioreg.h: revision 1.12
	sys/dev/pci/viornd.c: revision 1.19
	sys/dev/pci/virtio.c: revision 1.66
	sys/dev/pci/virtio.c: revision 1.67
	sys/dev/pci/virtio.c: revision 1.68
	sys/dev/pci/if_vioif.c: revision 1.103
	sys/dev/pci/virtio.c: revision 1.69
	sys/dev/pci/if_vioif.c: revision 1.104
	sys/dev/pci/virtio_pci.c: revision 1.40
	sys/dev/virtio/virtio_mmio.c: revision 1.8
	sys/dev/virtio/virtio_mmio.c: revision 1.9
	sys/dev/pci/viomb.c: revision 1.14
	sys/dev/pci/viomb.c: revision 1.15
	sys/dev/pci/viomb.c: revision 1.17
	sys/dev/pci/viornd.c: revision 1.20
	sys/dev/pci/viornd.c: revision 1.21
	sys/dev/pci/virtiovar.h: revision 1.25
	sys/dev/pci/virtiovar.h: revision 1.26
	sys/dev/pci/virtiovar.h: revision 1.27
	sys/dev/pci/virtiovar.h: revision 1.28
	sys/dev/pci/virtio.c: revision 1.70
	sys/dev/pci/virtio.c: revision 1.71
	sys/dev/pci/virtio.c: revision 1.72
	sys/dev/pci/virtio.c: revision 1.73
	sys/dev/pci/virtio.c: revision 1.74
	sys/dev/pci/virtio_pci.c: revision 1.39

Set virtqueues in virtio_child_attach_finish

The number of virtqueue maybe change in a part of VirtIO devices
(e.g. vioif(4)). And it is fixed after negotiation of features.
So the configuration is moved into the function.

viocon(4): fix not to allocate unused virtqueue

viocon(4) allocates 4 virtqueues but it only uses 2 (0 and 1) queues.

Added functions to set interrupt handler and index into virtqueue

Added check of pointer for allocated memory before release of resource

Setup virtqueues after registering them to virtio_softc
restore fetch of qsize.

Mark as MPSAFE.

virtio(4): Avoid name collision with global intrhand on sparc64.

Pacifies -Werror=shadow.
No functional change intended.

Use PRIuBUSSIZE to print bus_size_t variables.

virtio(4): Fix sizing of virtqueue allocation.
vq->vq_avail[0].ring is a zero-length array, and thus sizeof is zero;
likewise vq->vq_used[0].ring.
Use vq->vq_avail[0].ring[0] and vq->vq_used[0].ring[0] to fix this
and restore the previous allocation sizing logic.
XXX We shouldn't use zero-length arrays here -- they are asking for
trouble like this, and C99 has a standard way to express what we're
actually trying to get at it, flexible array members.
PR kern/57304

virtio(4): Use flexible array members, not zero-length arrays.
This enables the compiler to detect sizeof mistakes like
PR kern/57304.

Use descriptor chain for free slots instead of vq_entry list
Descriptors can be chained by themself. And descriptors added to
avail ring or used ring are already chained. But it was not used
for unused descriptors and another linked list structure named
vq_entry was used.

The chain is also used for unused descriptors to make virtio(4)
simpler.

Added flags to store status of attaching a virtio device

This prevents a panic on reboot after a virtio device had called
virtio_child_attach_failed().

Fix wrong variable names
This fixes build errors in virtio_mmio.c

Mark as MPSAFE.

A single SCSI request may require multiple slots in the virtio queue
but the queue isn't sized for the theoretical maximum. So just use
XS_BUSY to pace the scsipi layer.

restore fetch of qsize.

Added check of pointer for allocated memory before release of resource

Added functions to set interrupt handler and index into virtqueue

Set virtqueues in virtio_child_attach_finish

The number of virtqueue maybe change in a part of VirtIO devices
(e.g. vioif(4)). And it is fixed after negotiation of features.
So the configuration is moved into the function.

fix typos in log messages s/bus_dmamem_create/bus_dmamap_create/ and
s/bus_dmamem_load/bus_dmamap_load/.
Inspired by recent similar fixes in OpenBSD.

We no need for PQUIRK_FORCELUNS now that we try REPORT LUNS.

Trailing whitespace

Merge thorpej-cfargs2.

Adapt to CFARGS().

Merge thorpej-cfargs branch:

Simplify and make extensible the config_search() / config_found() /
config_attach() interfaces: rather than having different variants for
which arguments you want pass along, just have a single call that
takes a variadic list of tag-value arguments.

Adjust all call sites:
- Simplify wherever possible; don't pass along arguments that aren't
  actually needed.
- Don't be explicit about what interface attribute is attaching if
  the device only has one.  (More simplification.)
- Add a config_probe() function to be used in indirect configuiration
  situations, making is visibly easier to see when indirect config is
  in play, and allowing for future change in semantics.  (As of now,
  this is just a wrapper around config_match(), but that is an
  implementation detail.)

Remove unnecessary or redundant interface attributes where they're not
needed.

There are currently 5 "cfargs" defined:
- CFARG_SUBMATCH (submatch function for direct config)
- CFARG_SEARCH (search function for indirect config)
- CFARG_IATTR (interface attribte)
- CFARG_LOCATORS (locators array)
- CFARG_DEVHANDLE (devhandle_t - wraps OFW, ACPI, etc. handles)

...and a sentinel value CFARG_EOL.

Add some extra sanity checking to ensure that interface attributes
aren't ambiguous.

Use CFARG_DEVHANDLE in MI FDT, OFW, and ACPI code, and macppc and shark
ports to associate those device handles with device_t instance.  This
will trickle trough to more places over time (need back-end for pre-OFW
Sun OBP; any others?).

Sync with HEAD.

Give config_found() the same variadic arguments treatment as
config_search().  This commit only adds the CFARG_EOL sentinel
to the existing config_found() calls.  Conversion of config_found_sm_loc()
and config_found_ia() call sites will be in subsequent commits.

Add VirtIO PCI v1.0 attachments and fix the drivers affected.

The vioif, ld, scsi, viornd and viomb devices were adjusted when needed and
tested both in legacy 0.9 and v1.0 attachments trough PCI on amd64, sparc64,
aarch64 and aarch64-eb. ACPI/FDT attachments also tested on
aarch64/aarch64-eb.

Known issues

* viomb on aarch64 works only with ACPI/FDT attachment but not with PCI
  attachment. PCI and ACPI/FDT attachment works on aarch64-eb.

* virtio on sparc64 attaches but is it not functioning though not a
  regression.

Update the Proxmox comment with link to their bugzilla

Probe all LUNs on vioscsi(4) until scsi(4) can do REPORT LUNS.

Pull up following revision(s) (requested by kim in ticket #1572):

	sys/dev/pci/vioscsi.c: revision 1.22

Fix off-by-one SCSI target reporting

Use max_target as the controller ID instead of zero, so that the device
located at SCSI ID zero (e.g. a disk) is not obscured through not being
probed by scsi_probe_bus() (which skips the controller ID).

Copy the target requested onto the wire without decrementing it by one.

ok christos@

Pull up following revision(s) (requested by kim in ticket #1001):

	sys/dev/pci/vioscsi.c: revision 1.22

Fix off-by-one SCSI target reporting

Use max_target as the controller ID instead of zero, so that the device
located at SCSI ID zero (e.g. a disk) is not obscured through not being
probed by scsi_probe_bus() (which skips the controller ID).

Copy the target requested onto the wire without decrementing it by one.

ok christos@

Fix off-by-one SCSI target reporting

Use max_target as the controller ID instead of zero, so that the device
located at SCSI ID zero (e.g. a disk) is not obscured through not being
probed by scsi_probe_bus() (which skips the controller ID).

Copy the target requested onto the wire without decrementing it by one.

ok christos@

Pull up following revision(s) (requested by maxv in ticket #1289):

	sys/dev/pci/vioscsi.c: revision 1.21

Fix use-after-free. If we're not polling, virtio_enqueue_commit() will send
the transaction, and it means 'xs' can be immediately freed. So, save the
value of xs_control beforehand.

Detected by KASAN, ok jdolecek@.
Fixes PR/54008

Sync with HEAD

Fix use-after-free. If we're not polling, virtio_enqueue_commit() will send
the transaction, and it means 'xs' can be immediately freed. So, save the
value of xs_control beforehand.

Detected by KASAN, ok jdolecek@.

Fixes PR/54008
Reported-by: syzbot+6513c4afe66237d7207f@syzkaller.appspotmail.com

Sync with HEAD

remove irrelevant pci(9) #includes from virtio child drivers

update from HEAD

Sync with HEAD

Resolve conflicts from previous merge (all resulting from $NetBSD
keywork expansion)

file vioscsi.c was added on branch tls-maxphys on 2017-12-03 11:37:29 +0000

free the virtio slot _before_ calling scsipi_done(), so that driver is able
to accept another request immediatelly and doesn't trigger spurious
XS_RESOURCE_SHORTAGE

part of fix for PR kern/52199

refactor error handling in vioscsi_scsipi_request() to avoid the goto maze,
and add missing virtio_enqueue_abort() call when bus_dmamap_load() fails;
abort is only done implicitely when virtio_enqueue_reserve() fails,
must do it explicitely if there is other reason

also now always print error when bus_dmamap_load() or virtio_enqueue_reserve()
fail - the former shouldn't fail due to BUS_DMA_ALLOCNOW, and the latter
shouldn't ever fail now after the maxnsegs fix for virtio_alloc_vq(), so
error there means driver bug

fix number of maximum segments in the queue allocation so that
MAXPHYS-sized tranfer fits inside the indirect descriptor; avoids the
!indirect path in virtio_enqueue_reserve() for these transfers,
which runs out of the vq_entries and triggers XS_RESOURCE_SHORTAGE for
bigger number of outstanding tranfers

fixes PR kern/52199 by MB

Sync with HEAD

Sync with HEAD

fix printf format.

add support for loading vioscsi driver dynamically

cap also adapt_nchannels to 1, we only attach one

cap adapt_openings to minimum of cmd_per_lun and qsize, so that we don't
overpromise to the upper layer

cap ntargets and nluns to 16 and 1024 as avoid spending excessive time
probing for nonexistent devices; e.g. qemu defaults to 255/16383 causing
the probe to take about a minute

reorganize the attachment process for virtio child devices, so that
more common code is shared among the drivers, and it's possible for
the drivers to be correctly dynamically loaded; forbid direct access
to struct virtio_softc from the child driver code

Sync with HEAD

several small tweaks:
- use MSIX if available
- fix off-by-one for nluns/ntargets (code inspection, not actual effect)
- when virtio_enqueue_reserve() fails, do not execute vioscsi_req_put()
  and hence virtio_dequeue_commit() as there is nothing to commit, just
  unload the xs and return; also report XS_RESOURCE_SHORTAGE, it's
  semi-normal situation
- set status/resid in vioscsi_req_done() before the switch, so that
  the override for VIRTIO_SCSI_S_BAD_TARGET actually has effect
- g/c vioscsi_req_put(), call the appropriate cleanup routines directly
- stop initializing id/task_attr in vioscsi_req_get(), it's overwritten
  in vioscsi_scsipi_request() anyway

replace req queue offset with define, only set & start the req interrupt for
that queue; add some masks for the lun for visibility; use slot as tag number

no actual functional change, just small cleanup

allocate bus dma maps during attachment, rather than creating and destroying
them for each request; besides being faster, bus_dmamap_destroy() is not
safe to be called from interrupt context

adresses PR kern/52034 by Benny Siegert

Sync with netbsd-5

Pull up following revision(s) (requested by bsiegert in ticket #1338):
	sys/arch/amd64/conf/GENERIC: revision 1.418
	sys/dev/pci/files.pci: revision 1.379
	sys/dev/pci/vioscsi.c: revisions 1.1-1.8
	sys/dev/pci/vioscsireg.h: revision 1.1
	sys/dev/pci/virtioreg.h: revisions 1.5, 1.6
add the common flag bits
--
Add vioscsi, compile tested only (toxic)
--
this is working now, remove debugging.
--
add vioscsi
--
kill some more debugging.
--
Fake mode sense data for illegal targets. qemu reports 256 targets...
--
fix the status return.
--
Use SCSIPI_CHAN_NOSETTLE.
--
vioscsi_req_get()/virtio_enqueue_prep() failing is actually perfectly
normal - observed failures included 10, 27, 61 in-flight commands,
so probably depends on particular command mix; return with
XS_RESOURCE_SHORTAGE rather then panic
do vioscsi_req_put() when initial bus_dmamap_load() fails, as suggested
by the XXX; the vq_done hook is called by virtio, but in that case we
never get to commit the request to it
--
react on ADAPTER_REQ_SET_XFER_MODE so that we set tagged queuing
pass tag type and set id

Sync with HEAD

Sync with HEAD

file vioscsi.c was added on branch netbsd-7-nhusb on 2017-01-18 08:46:42 +0000

file vioscsi.c was added on branch netbsd-7 on 2016-12-23 06:33:56 +0000

react on ADAPTER_REQ_SET_XFER_MODE so that we set tagged queuing

pass tag type and set id

vioscsi_req_get()/virtio_enqueue_prep() failing is actually perfectly
normal - observed failures included 10, 27, 61 in-flight commands,
so probably depends on particular command mix; return with
XS_RESOURCE_SHORTAGE rather then panic

do vioscsi_req_put() when initial bus_dmamap_load() fails, as suggested
by the XXX; the vq_done hook is called by virtio, but in that case we never
get to commit the request to it

Sync with HEAD (as of 26th Dec)

file vioscsi.c was added on branch nick-nhusb on 2015-12-27 12:09:57 +0000

Use SCSIPI_CHAN_NOSETTLE.

fix the status return.

Fake mode sense data for illegal targets. qemu reports 256 targets...

kill some more debugging.

this is working now, remove debugging.

Add vioscsi, compile tested only (toxic)