Since PDMA reads/writes are driven by the guest, it is possible that migration
can occur whilst a SCSIRequest is still active. Fortunately active SCSIRequests
are already included in the migration stream and restarted post migration but
this still leaves the reference in ESPState uninitialised.
Implement the SCSIBusInfo .load_request callback to obtain a reference to the
currently active SCSIRequest and use it to recreate ESPState current_req
after migration.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20220305155530.9265-11-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
This involves (re)adding a PDMA-specific subsection to hold the reference to the
current PDMA callback.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20220305155530.9265-10-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
This prepares for the inclusion of the current PDMA callback in the migration
stream since the callback is referenced by an integer instead of a function
pointer.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20220305155530.9265-9-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
This function is to be used to execute the current PDMA callback rather than
dereferencing the ESPState pdma_cb function pointer directly.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20220305155530.9265-8-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
This function is to be used to set the current PDMA callback rather than
accessing the ESPState pdma_cb function pointer directly.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20220305155530.9265-7-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
If a reset command is sent after data has been transferred into the SCSI buffer
ensure that async_len is reset to 0. Otherwise a subsequent TI command assumes
the SCSI buffer contains data to be transferred to the device causing it to
dereference the stale async_buf pointer.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Fixes: https://gitlab.com/qemu-project/qemu/-/issues/724
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20211118100327.29061-2-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
There is currently a check in esp_select() to cancel any in-flight SCSI requests
to ensure that issuing multiple select commands without continuing through the
rest of the ESP state machine ignores all but the last SCSI request. This is
also enforced through the addition of assert()s in esp_transfer_data() and
scsi_read_data().
The get_cmd() function does not call esp_select() when TC == 0 which means it is
possible for a fuzzer to trigger these assert()s by sending a select command when
TC == 0 immediately after a valid SCSI CDB has been submitted.
Since esp_select() is only called from get_cmd(), hoist the check to cancel
in-flight SCSI requests from esp_select() into get_cmd() to ensure it is always
called when executing a select command to initiate a new SCSI request.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Closes: https://gitlab.com/qemu-project/qemu/-/issues/662
Closes: https://gitlab.com/qemu-project/qemu/-/issues/663
Message-Id: <20211101183516.8455-2-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The function scsi_bus_new() creates a new SCSI bus; callers can
either pass in a name argument to specify the name of the new bus, or
they can pass in NULL to allow the bus to be given an automatically
generated unique name. Almost all callers want to use the
autogenerated name; the only exception is the virtio-scsi device.
Taking a name argument that should almost always be NULL is an
easy-to-misuse API design -- it encourages callers to think perhaps
they should pass in some standard name like "scsi" or "scsi-bus". We
don't do this anywhere for SCSI, but we do (incorrectly) do it for
other bus types such as i2c.
The function name also implies that it will return a newly allocated
object, when it in fact does in-place allocation. We more commonly
name such functions foo_init(), with foo_new() being the
allocate-and-return variant.
Replace all the scsi_bus_new() callsites with either:
* scsi_bus_init() for the usual case where the caller wants
an autogenerated bus name
* scsi_bus_init_named() for the rare case where the caller
needs to specify the bus name
and document that for the _named() version it's then the caller's
responsibility to think about uniqueness of bus names.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20210923121153.23754-2-peter.maydell@linaro.org
The LUN is selected with an IDENTIFY message, and persists
until the next message out phase. Instead of passing it to
do_busid_cmd, store it in ESPState. Because do_cmd can simply
skip the message out phase if cmdfifo_cdb_offset is zero, it
can now be used for the S without ATN cases as well.
Reviewed-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Tested-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit 4e78f3bf35 "esp: defer command completion interrupt on incoming data
transfers" added a version check for use with VMSTATE_*_TEST macros to allow
migration from older QEMU versions. Unfortunately the version check fails to
work in its current form since if the VMStateDescription version_id is
incremented, the test returns false and so the fields are not included in the
outgoing migration stream.
Change the version check to use >= rather == to ensure that migration works
correctly when the ESPState VMStateDescription has version_id > 5.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Fixes: 4e78f3bf35 ("esp: defer command completion interrupt on incoming data transfers")
Message-Id: <20210613102614.5438-1-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit 799d90d818 "esp: transition to message out phase after SATN and stop
command" added logic to correctly handle extended messages for DMA requests
but not for PDMA requests.
Apply the same logic in esp_do_dma() to do_dma_pdma_cb() so that extended
messages terminated with a PDMA request are accumulated correctly. This allows
the ESP device to respond correctly to the SDTR negotiation initiated by the
NetBSD ESP driver without causing errors and timeouts on boot.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20210519100803.10293-6-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This commit from nearly 10 years ago is now broken due to the improvements
in esp emulation (or perhaps was never correct). It shows up as a bug
in detecting the CDROM drive under MacOS. The error is caused by the
MacOS CDROM driver sending this CDB with an "S without ATN" command and
without DMA:
0x12 0x00 0x00 0x00 0x05 0x00 (INQUIRY)
This is a valid INQUIRY command, however with this logic present the 3rd
byte (0x0) is copied over the 1st byte (0x12) which silently converts the
INQUIRY command to a TEST UNIT READY command before passing it to the
QEMU SCSI layer. Since the TEST UNIT READY command has a zero length
response the MacOS CDROM driver never receives a response and assumes
the CDROM is not present.
The logic was to ignore the IDENTIFY byte and copy the LUN over from
the CDB, which did store the LUN in bits 5-7 of the second byte in
olden times. This however is all obsolete, so just drop the code.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20210519100803.10293-5-mark.cave-ayland@ilande.co.uk>
[Tweaked commit message. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
After each PDMA write transfer the MacOS CDROM driver waits until the FIFO is empty
(i.e. its contents have been written out to the SCSI bus) by polling the FIFO count
register until it reads 0. This doesn't work with the current PDMA write
implementation which waits until either the FIFO is full or the transfer is complete
before invoking the PDMA callback to process the FIFO contents.
Change the PDMA write transfer logic so that the PDMA callback is invoked after each
PDMA write to transfer the FIFO contents to the target buffer immediately, and hence
avoid getting stuck in the FIFO count register polling loop.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20210519100803.10293-4-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The initial implementation of non-DMA transfers was based upon analysis of traces
from the MacOS toolbox ROM for handling unaligned reads but missed one key
aspect - during a non-DMA transfer from the target, the bus service interrupt
should be raised for every single byte received from the bus and not just at either
the end of the transfer or when the FIFO is full.
Adjust the non-DMA code accordingly so that esp_do_nodma() is called for every byte
received from the target. This also includes special handling for managing the change
from DATA IN to STATUS phase as this needs to occur when the final byte is read out
from the FIFO, and not at the end of the transfer of the last byte into the FIFO.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20210519100803.10293-3-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The current implementation only resumes DMA transfers when incoming data is
received from the target device, but this is also required for non-DMA transfers
with the next set of non-DMA changes.
Rather than duplicate the DMA/non-DMA dispatch logic in the initial transfer
section, update the code so that the initial transfer section can just
fallthrough to the main DMA/non-DMA dispatch logic.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Message-Id: <20210519100803.10293-2-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When processing a command to select a target and send a CDB, the ESP device
maintains a sequence step register so that if an error occurs the host can
determine which part of the selection/CDB submission sequence failed.
The old Linux 2.6 driver is really pedantic here: it checks the sequence step
register even if a command succeeds and complains loudly on the console if the
sequence step register doesn't match the expected bus phase and interrupt flags.
This reason this mismatch occurs is because the ESP emulation currently doesn't
update the bus phase until the next TI (Transfer Information) command and so the
cleared sequence step register is considered invalid for the stale bus phase.
Normally this isn't an issue as the host only checks the sequence step register
if an error occurs but the old Linux 2.6 driver does this in several places
causing a large stream of "esp0: STEP_ASEL for tgt 0" messages to appear on the
console during the boot process.
Fix this by not clearing the sequence step register when reading the interrupt
register and clearing the DMA status, so the guest sees a valid sequence step
and bus phase combination at the end of the command phase. No other change is
required since the sequence step register is correctly updated throughout the
selection/CDB submission sequence once one of the select commands is issued.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Fixes: 1b9e48a5bd ("esp: implement non-DMA transfers in PDMA mode")
Message-Id: <20210518212511.21688-3-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The datasheet sequence tables confirm that when a target selection fails, only
the INTR_DC interrupt flag should be asserted.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Fixes: cf47a41e05 ("esp: latch individual bits in ESP_RINTR register")
Message-Id: <20210518212511.21688-2-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When a CDB has been received and is about to be submitted to the SCSI layer
via one of the ESP select commands, ensure that do_cmd is set to zero before
executing the command.
Otherwise a guest executing 2 valid CDBs in quick sequence can invoke the SCSI
.transfer_data callback again before do_cmd is set to zero by the callback
function triggering an assert at the start of esp_transfer_data().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210407195801.685-12-mark.cave-ayland@ilande.co.uk>
Instead let the SCSI layer invoke the .cancel callback itself to cancel and
reset the request state.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210407195801.685-11-mark.cave-ayland@ilande.co.uk>
If a guest transfers the message out/command phase data using DMA with a TC
that is larger than the cmdfifo size then the cmdfifo overflows triggering
an assert. Limit the size of the transfer to the free space available in
cmdfifo.
Buglink: https://bugs.launchpad.net/qemu/+bug/1919036
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-10-mark.cave-ayland@ilande.co.uk>
If the guest tries to read a CDB using DMA and cmdfifo is not empty then it is
possible to overflow cmdfifo.
Since this can only occur by issuing deliberately incorrect instruction
sequences, ensure that the maximum length of the CDB transferred to cmdfifo is
limited to the available free space within cmdfifo.
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-9-mark.cave-ayland@ilande.co.uk>
If the guest tries to execute a CDB when cmdfifo is not empty before the start
of the message out phase then clearing the message out phase data will cause
cmdfifo to underflow due to cmdfifo_cdb_offset being larger than the amount of
data within.
Since this can only occur by issuing deliberately incorrect instruction
sequences, ensure that the maximum length of esp_fifo_pop_buf() is limited to
the size of the data within cmdfifo.
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-8-mark.cave-ayland@ilande.co.uk>
When about to execute a SCSI command, ensure that cmdfifo is not empty and
current_dev is non-NULL. This can happen if the guest tries to execute a TI
(Transfer Information) command without issuing one of the select commands
first.
Buglink: https://bugs.launchpad.net/qemu/+bug/1910723
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-7-mark.cave-ayland@ilande.co.uk>
The const pointer returned by fifo8_pop_buf() lies directly within the array used
to model the FIFO. Building with address sanitizers enabled shows that if the
caller expects a minimum number of bytes present then if the FIFO is nearly full,
the caller may unexpectedly access past the end of the array.
Introduce esp_fifo_pop_buf() which takes a destination buffer and performs a
memcpy() in it to guarantee that the caller cannot overwrite the FIFO array and
update all callers to use it. Similarly add underflow protection similar to
esp_fifo_push() and esp_fifo_pop() so that instead of triggering an assert()
the operation becomes a no-op.
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20210407195801.685-6-mark.cave-ayland@ilande.co.uk>
Each FIFO currently has its own pop functions with the only difference being
the capacity check. The original reason for this was that the fifo8
implementation doesn't have a formal API for retrieving the FIFO capacity,
however there are multiple examples within QEMU where the capacity field is
accessed directly.
Change esp_fifo_pop() to access the FIFO capacity directly and then consolidate
esp_cmdfifo_pop() into esp_fifo_pop().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-5-mark.cave-ayland@ilande.co.uk>
Each FIFO currently has its own push functions with the only difference being
the capacity check. The original reason for this was that the fifo8
implementation doesn't have a formal API for retrieving the FIFO capacity,
however there are multiple examples within QEMU where the capacity field is
accessed directly.
Change esp_fifo_push() to access the FIFO capacity directly and then consolidate
esp_cmdfifo_push() into esp_fifo_push().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-4-mark.cave-ayland@ilande.co.uk>
The code for write_response() has always used the FIFO to store the data for
the status/message in phases, even for DMA transactions. Switch to using a
separate buffer that can be used directly for DMA transactions and restrict
the FIFO use to the non-DMA case.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210407195801.685-3-mark.cave-ayland@ilande.co.uk>
After issuing a SCSI command the SCSI layer can call the SCSIBusInfo .cancel
callback which resets both current_req and current_dev to NULL. If any data
is left in the transfer buffer (async_len != 0) then the next TI (Transfer
Information) command will attempt to reference the NULL pointer causing a
segfault.
Buglink: https://bugs.launchpad.net/qemu/+bug/1910723
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-2-mark.cave-ayland@ilande.co.uk>
If QEMU is launched with the -S option then the ESPState mig_version_id property
is left unset due to the ordering of the VMState fields in the VMStateDescription
for sysbusespscsi and pciespscsi. If the VM is migrated and restored in this
stopped state, the version tests in the vmstate_esp VMStateDescription and
esp_post_load() become confused causing the migration to fail.
Fix the ordering problem by moving the setting of mig_version_id to a common
esp_pre_save() function which is invoked first by both sysbusespscsi and
pciespscsi rather than at the point where ESPState is itself serialised into the
migration stream.
Buglink: https://bugs.launchpad.net/qemu/+bug/1922611
Fixes: 0bd005be78 ("esp: add vmstate_esp version to embedded ESPState")
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20210407124842.32695-1-mark.cave-ayland@ilande.co.uk>
When the MacOS toolbox ROM transfers data from a target device to an unaligned
memory address, the first/last byte of a 16-bit transfer needs to be handled
separately. This means that the first byte is preloaded into the FIFO before
the transfer, or the last byte remains in the FIFO after the transfer.
The result of this is that the PDMA routines must be updated so that the FIFO
is loaded/unloaded if the last 16-bit word is used (rather than the last byte)
and any remaining byte from a FIFO wraparound is handled correctly.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-43-mark.cave-ayland@ilande.co.uk>
The MacOS toolbox ROM uses non-DMA TI commands to handle the first/last byte
of an unaligned 16-bit transfer to memory.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-42-mark.cave-ayland@ilande.co.uk>
The bottom 5 bits contain the number of bytes remaining in the FIFO which is
trivial to implement with Fifo8 (the remaining bits are unimplemented and left
as 0 for now).
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-41-mark.cave-ayland@ilande.co.uk>
Rename ESP_CMDBUF_SZ to ESP_CMDFIFO_SZ and cmdbuf_cdb_offset to cmdfifo_cdb_offset
to indicate that the command buffer type has changed from an array to a Fifo8.
This also enables us to remove the ESPState field cmdlen since the command length
is now simply the number of elements used in cmdfifo.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-40-mark.cave-ayland@ilande.co.uk>
Rename TI_BUFSZ to ESP_FIFO_SZ since this constant is really describing the size
of the FIFO and is not directly related to the TI size.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-39-mark.cave-ayland@ilande.co.uk>
The SCSI bus should remain in the message out phase after the SATN and stop
command rather than transitioning to the command phase. A new ESPState variable
cmdbuf_cdb_offset is added which stores the offset of the CDB from the start
of cmdbuf when accumulating extended message out phase data.
Currently any extended message out data is discarded in do_cmd() before the CDB
is processed in do_busid_cmd().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-38-mark.cave-ayland@ilande.co.uk>
Some guests use a mixture of DMA and non-DMA transfers in combination with the
SATN and stop command to transfer message out phase and command phase bytes to
the target. Prepare for the next commit by adding a maxlen parameter to
get_cmd() to allow partial transfers.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-37-mark.cave-ayland@ilande.co.uk>
This matches the description in the datasheet and is required as support for
non-DMA transfers is added.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-36-mark.cave-ayland@ilande.co.uk>
Commit ea84a44250 "scsi: esp: Defer command completion until previous interrupts
have been handled" provided a mechanism to delay the command completion interrupt
until ESP_RINTR is read after the command has completed.
With the previous fixes for latching the ESP_RINTR bits and deferring the setting
of the command completion interrupt for incoming data to the SCSI callback, this
workaround is no longer required and can be removed.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-35-mark.cave-ayland@ilande.co.uk>
The MacOS toolbox ROM issues a command to the ESP controller as part of its
"FAST" SCSI routines and then proceeds to read the incoming data soon after
receiving the command completion interrupt.
Unfortunately due to SCSI block transfers being asynchronous the incoming data
may not yet be present causing an underflow error. Resolve this by waiting for
the SCSI subsystem transfer_data callback before raising the command completion
interrupt.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-34-mark.cave-ayland@ilande.co.uk>
Currently the ESP_RINTR register is set to a specific value as required within
the ESP state machine. In order to implement the upcoming deferred interrupt
functionality it is necessary to set individual bits within ESP_RINTR so that
a deferred interrupt will not overwrite the value of any other interrupt bits.
This also requires fixing up a few locations where the ESP_RINTR and ESP_RSEQ
registers are set/reset unexpectedly.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-33-mark.cave-ayland@ilande.co.uk>
At this point it is now possible to properly implement the FIFO flush command
without causing guest errors.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-32-mark.cave-ayland@ilande.co.uk>
The MacOS toolbox ROM performs 4 byte reads/writes when transferring data to
and from the target. Since the SCSI bus is 16-bits wide, use the memory API
to split a 4 byte access into 2 x 2 byte accesses.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-31-mark.cave-ayland@ilande.co.uk>
Now that all data is transferred via the FIFO (ti_buf) there is no need to track
the source buffer being used for the data transfer. This also eliminates the
need for a separate subsection for PDMA state migration.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-30-mark.cave-ayland@ilande.co.uk>
PDMA as implemented on the Quadra 800 uses DREQ to load data into the FIFO
up to a maximum of 16 bytes at a time. The MacOS toolbox ROM requires this
because it mixes FIFO and PDMA transfers whilst checking the FIFO status
and counter registers to ensure success.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-29-mark.cave-ayland@ilande.co.uk>
Currently the target selection for PDMA is done after the SCSI command has been
delivered which is not correct. Perform target selection as part of the initial
get_cmd() call when the command is submitted: if no target is present, don't
raise DRQ.
If the target is present then switch to the command phase since the MacOS toolbox
ROM checks for this before attempting to submit the SCSI command.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-28-mark.cave-ayland@ilande.co.uk>
This better describes the purpose of the function.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-27-mark.cave-ayland@ilande.co.uk>
The cmdbuf is really just a copy of FIFO data (including extra message phase
bytes) so its pdma_origin is effectively TI. Fortunately we already know when
we are receiving a SCSI command since do_cmd == 1 which enables us to
distinguish between the two cases in esp_pdma_read()/esp_pdma_write().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-26-mark.cave-ayland@ilande.co.uk>
Real hardware simply counts down using the in-built TC to determine when the
the PDMA request is complete. Use the TC to determine the PDMA transfer length
which then enables us to remove the redundant pdma_len variable.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-25-mark.cave-ayland@ilande.co.uk>
This eliminates the last user of the PDMA-specific pdma_cur variable which can
now be removed.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-24-mark.cave-ayland@ilande.co.uk>
Here the updates to async_len and ti_size are moved into the corresponding
esp_pdma_read()/esp_pdma_write() function to eliminate the reference to
pdma_cur in do_dma_pdma_cb().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210304221103.6369-23-mark.cave-ayland@ilande.co.uk>
