Since there's no support for exception handling in C, the try-catch
blocks have been deleted, and throw clauses are replaced. When a runtime
error happens, we're printing out the error message. Disassembling of
the current instruction interrupts. This behavior is achieved by adding
sigsetjmp() to discard further disassembling after the error message
prints and by adding the siglongjmp() function to imitate throwing an
error. The goal was to maintain the same output as it was.
This patch expands the Dis_info struct, which should hold the
necessary data for handling runtime errors. Fields fprintf_func and
stream are in charge of error printing. Field buf enables the use of
sigsetjmp() and siglongjmp() functions. Support for runtime error
handling will be added later.
We're filling Dis_info at the entrance of the nanoMIPS disassembler,
i.e. print_insn_nanomips. Next, we're adding that information as an
argument wherever we need to.
Before this patch, we'd been calling img_format twice, the first time
through the CPR function to get an appropriate string and the second
time to print that formatted string. There's no more need for that.
Therefore, calls to CPR are removed, and now we're directly printing
"CP" and integer value instead.
Both versions of IMMEDIATE functions have been removed.
Before this patch, we'd been calling img_format twice, the first time
through the IMMEDIATE to get an appropriate string and the second time
to print that string. There's no more need for that. Therefore, calls to
IMMEDIATE are removed, and now we're directly printing the integer
values instead.
The return type of typedef disassembly_function is changed to char *
instead of std::string. Therefore, for every particular
disassembly_function function signature is changed.
For example:
- static std::string ABS_D(uint64 instruction, img_address m_pc) {...}
is replaced with
- static char *ABS_D(uint64 instruction, img_address m_pc) {...}
Every helper function used to return std::string is changed to return
const char * or char *. Where the return value points to a static string
that the caller must not free, the return type is const char *. If a
function allocates memory and the caller is required to free it, the
return type is a char *. This applies to the following functions:
img_format, to_string, GPR, save_restore_list, FPR, etc.
Now that we replaced every std::string for const char * or char *, it is
possible to delete multiple versions of the img_format function. The
general version:
- static char *img_format(const char *format, ...) {...}
can handle all string formatting, so others have been deleted.
Where necessary, strings are dynamically allocated with g_strjoinv,
g_strdup, g_strdup_vprintf, and g_strdup_printf. Memory leaking will be
prevented later.
String concatenation in the save_restore_list() function is handled
using g_strjoinv() function instead of += operator.
The type of the "dis" parameter in the Disassemble function is changed
- from std::string &
- to char **
Without applying all of these changes, the nanomips disassembler may be
buildable but can't produce the appropriate output, so all of them are
made together.
Following functions just wrap the decode_gpr_gpr3() function:
- encode_rs3_and_check_rs3_ge_rt3()
- encode_rs3_and_check_rs3_lt_rt3()
Therefore those have been deleted. Calls to these two functions have
been replaced with calls to decode_gpr_gpr3.
Header file nanomips.h has been deleted for the nanomips disassembler to
stay consistent with the rest of the disassemblers which don't include
extra header files.
The following is moved from the nanomips.h to nanomips.cpp file:
- #include line
- typedefs
- enums
- definition of the Pool struct.
Header file nanomips.h will be deleted to be consistent with the rest of
the disas/ code.
NMD class has been deleted. The following methods are now declared as
static functions:
- public NMD::Disassemble method
- private NMD::Disassemble method
- private NMD::extract_op_code_value helper method
Also, the implementation of the print_insn_nanomips function and
nanomips_dis function is moved to the end of the nanomips.cpp file,
right after the implementation of the Disassemble function.
Pool tables are no longer declared as static fields of the NMD
class but as global static const variables. Pool struct is defined
outside of the class.
The NMD::Disassemble method is using the MAJOR Pool table variable, so
its implementation is moved to the end of the nanomips.cpp file,
right after the initialization of the MAJOR Pool table.
NMD class methods with the disassembly_function type like
NMD::ABS_D, NMD::ABS_S, etc. are removed from the class. They're now
declared global static functions. Therefore, typedef of the function
pointer, disassembly_function is defined outside of the class.
Now that disassembly_function type functions are not part of the NMD
class we can't access them using the this pointer. Thus, the use of
the this pointer has been deleted.
NMD class methods with the conditional_function type like
NMD::ADDIU_32__cond, NMD::ADDIU_RS5__cond, etc. are removed from the NMD
class. They're now declared global static functions. Therefore, typedef
of the function pointer, conditional_function is defined outside of the
class.
Now that conditional_function type functions are not part of the NMD
class we can't access them using the this pointer. Thus, the use of
the this pointer has been deleted.
Helper methods from NMD class like NMD::renumber_registers,
NMD::decode_gpr_gpr4... etc. are removed from the class. They're now
declared global static functions.
Following helper methods have been deleted because they're not used by
the nanomips disassembler:
- NMD::encode_msbd_from_pos_and_size,
- NMD::encode_s_from_s_hi,
- NMD::neg_copy
Global functions used by those methods:
- nanomips_dis
- sign_extend
- extract_bits
have also been defined as static global functions.
We're deleting the m_pc field of the NMD class. It's now part of the
Dis_info struct that this patch introduces. Currently, the Dis_info
struct has just one field, m_pc, which we need for address calculation
in the ADDRESS function.
We're filling Dis_info at the entrance of the nanoMIPS disassembler.
I.e. print_insn_nanomips. Next, we're adding that information as an
argument wherever we need to.
Since NMD class now has no more fields, the NMD constructor is
also deleted.
The m_requested_instruction_categories field always has the same value,
ALL_ATTRIBUTES. The only use of that field is within the if statement.
When replaced with a specific value, the if statement is always false,
so it has been removed.
Now, when the only use of the m_requested_instruction_categories field
is removed, we can delete the field declaration and initialization in
the NMD class. Also, we're changing the way of the construction of the
NMD object in the nanomips_dis function.
Since there's no namespace feature in C, namespace img has been replaced
with adding the prefix "img" to the namespace members.
Prefix "img" has been added to the function names of functions that used
to be wrapped in namespace img. Those are img::format() functions.
I.e. replaced img::format with the img_format.
Typedef address that used to belong to namespace img now is called
img_address.
hw/isa/vt82c686: Create rtc-time alias in boards instead
According to good QOM practice, an object should only deal with objects
of its own sub tree. Having devices create an alias on the machine
object doesn't respect this good practice. To resolve this, create the
alias in the machine's code.
Bernhard Beschow [Mon, 13 Jun 2022 17:24:55 +0000 (19:24 +0200)]
hw/isa/vt82c686: Instantiate IDE function in host device
The IDE function is closely tied to the ISA function (e.g. the IDE
interrupt routing happens there), so it makes sense that the IDE
function is instantiated within the south bridge itself.
The object creation now happens in chip-specific init methods which
allows the realize methods to be consolidated into one method. Shifting
the logic into the init methods has the addidional advantage that the
parent object's init methods are called implicitly - like constructors
in object-oriented languages.
Stefan Hajnoczi [Sun, 30 Oct 2022 22:31:59 +0000 (18:31 -0400)]
Merge tag 'mem-2022-10-28' of https://github.com/davidhildenbrand/qemu into staging
Hi,
"Host Memory Backends" and "Memory devices" queue ("mem"):
- Fix NVDIMM error message
- Add ThreadContext user-creatable object and wire it up for NUMA-aware
hostmem preallocation
# -----BEGIN PGP SIGNATURE-----
#
# iQJFBAABCAAvFiEEG9nKrXNcTDpGDfzKTd4Q9wD/g1oFAmNbpHARHGRhdmlkQHJl
# ZGhhdC5jb20ACgkQTd4Q9wD/g1pDpw//bG9cyIlzTzDnU5pbQiXyLm0nF9tW/tli
# npGPSbFFYz/72XD9VJSVLhbNHoQSmFcMK5m/DA4WAMdOc5zF7lP3XdZcj72pDyxu
# 31hJRvuRhxNb09jhEdWRfX5+Jg9UyYXuIvtKXHSWgrtaYDtHBdTXq/ojZlvlo/rr
# 36v0jaVaTNRs7dKQL2oaN+DSMiPXHxBzA6FABqYmJNNwuMJT0kkX8pfz0OFwkRn+
# iqf9uRhM6b/fNNB0+ReA7FfGL+hzU6Uv8AvAL3orXUqjwPMRe9Fz2gE7HpFnE6DD
# dOP4Xk2iSSJ5XQA8HwtvrQfrGPh4gPYE80ziK/+8boy3alVeGYbYbvWVtdsNju41
# Cq9kM1wDyjZf6SSUIAbjOrNPdbhwyK4GviVBR1zh+/gA3uF5MhrDtZh4h3mWX2if
# ijmT9mfte4NwF3K1MvckAl7IHRb8nxmr7wjjhJ26JwpD+76lfAcmXC2YOlFGHCMi
# 028mjvThf3HW7BD2LjlQSX4UkHmM2vUBrgMGQKyeMham1VmMfSK32wzvUNfF7xSz
# o9k0loBh7unGcUsv3EbqUGswV5F6AgjK3vWRkDql8dNrdIoapDfaejPCd58kVM98
# 5N/aEoha4bAeJ6NGIKzD+4saiMxUqJ0y2NjSrE8iO4HszXgZW5e1Gbkn4Ae6d37D
# QSSqyfasVHY=
# =bLuc
# -----END PGP SIGNATURE-----
# gpg: Signature made Fri 28 Oct 2022 05:44:16 EDT
# gpg: using RSA key 1BD9CAAD735C4C3A460DFCCA4DDE10F700FF835A
# gpg: issuer "[email protected]"
# gpg: Good signature from "David Hildenbrand <[email protected]>" [unknown]
# gpg: aka "David Hildenbrand <[email protected]>" [full]
# gpg: aka "David Hildenbrand <[email protected]>" [unknown]
# gpg: WARNING: The key's User ID is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 1BD9 CAAD 735C 4C3A 460D FCCA 4DDE 10F7 00FF 835A
* tag 'mem-2022-10-28' of https://github.com/davidhildenbrand/qemu:
vl: Allow ThreadContext objects to be created before the sandbox option
hostmem: Allow for specifying a ThreadContext for preallocation
util: Make qemu_prealloc_mem() optionally consume a ThreadContext
util: Add write-only "node-affinity" property for ThreadContext
util: Introduce ThreadContext user-creatable object
util: Introduce qemu_thread_set_affinity() and qemu_thread_get_affinity()
util: Cleanup and rename os_mem_prealloc()
hw/mem/nvdimm: fix error message for 'unarmed' flag
* tag 'pull-error-2022-10-28' of https://repo.or.cz/qemu/armbru:
qerror: QERR_PERMISSION_DENIED is no longer used, drop
qtest: Improve error messages when property can not be set right now
backends: Improve error messages when property can no longer be set
qom: Improve error messages when property has no getter or setter
Stefan Hajnoczi [Sun, 30 Oct 2022 22:30:00 +0000 (18:30 -0400)]
Merge tag 'net-pull-request' of https://github.com/jasowang/qemu into staging
# -----BEGIN PGP SIGNATURE-----
# Version: GnuPG v1
#
# iQEcBAABAgAGBQJjW2i1AAoJEO8Ells5jWIR5HMIAIvDEmWQ2eZ1R+CfsefXkD5H
# W3RSZbMrOHR6sb9cbYpqK/vWmH8E/jZkKY4n/q7vQ3QerFMeDPgxu0Qn43iElLXS
# iGHhC51fa5IwJNDomjUGI8oJzyk0sxAbgwOjXZ4qbAkk9KeQYWU+JqYghvOrBYbd
# VaIwEiBlECuBy0DKx2gBTfxgeuw3V3WvtjpKeZVRlR+4vLQtI9Goga78qIPfjpH2
# sN/lFhZGaX1FT8DSft0oCCBxCK8ZaNzmMpmr39a8+e4g/EJZC9xbgNkl3fN0yYa5
# P0nluv/Z6e1TZ4FBDaiVFysTS5WnS0KRjUrodzctiGECE3sQiAvkWbKZ+QdGrlw=
# =0/b/
# -----END PGP SIGNATURE-----
# gpg: Signature made Fri 28 Oct 2022 01:29:25 EDT
# gpg: using RSA key EF04965B398D6211
# gpg: Good signature from "Jason Wang (Jason Wang on RedHat) <[email protected]>" [full]
# Primary key fingerprint: 215D 46F4 8246 689E C77F 3562 EF04 965B 398D 6211
* tag 'net-pull-request' of https://github.com/jasowang/qemu: (26 commits)
net: stream: add QAPI events to report connection state
net: stream: move to QIO to enable additional parameters
qemu-sockets: update socket_uri() and socket_parse() to be consistent
qemu-sockets: move and rename SocketAddress_to_str()
net: dgram: add unix socket
net: dgram: move mcast specific code from net_socket_fd_init_dgram()
net: dgram: make dgram_dst generic
net: stream: add unix socket
net: stream: Don't ignore EINVAL on netdev socket connection
net: socket: Don't ignore EINVAL on netdev socket connection
qapi: net: add stream and dgram netdevs
net: introduce qemu_set_info_str() function
qapi: net: introduce a way to bypass qemu_opts_parse_noisily()
net: simplify net_client_parse() error management
net: remove the @errp argument of net_client_inits()
net: introduce convert_host_port()
vhost: Accept event idx flag
vhost: use avail event idx on vhost_svq_kick
vhost: toggle device callbacks using used event idx
vhost: allocate event_idx fields on vring
...
Stefan Hajnoczi [Sun, 30 Oct 2022 19:15:12 +0000 (15:15 -0400)]
Merge tag 'for-upstream' of https://repo.or.cz/qemu/kevin into staging
Block layer patches
- Cleanup bs->backing and bs->file handling
- Refactor bdrv_try_set_aio_context using transactions
- Changes for improved coroutine_fn consistency
- vhost-user-blk: fix the resize crash
- io_uring: Use of io_uring_register_ring_fd() led to breakage, revert
- vvfat: Fix some problems with r/w mode
- Code cleanup
- MAINTAINERS: Fold "Block QAPI, monitor, ..." into "Block layer core"
* tag 'misc-next-pull-request' of https://gitlab.com/berrange/qemu:
crypto: add test cases for many malformed LUKS header scenarios
crypto: ensure LUKS tests run with GNUTLS crypto provider
crypto: quote algorithm names in error messages
crypto: split off helpers for converting LUKS header endianess
crypto: split LUKS header definitions off into file
crypto: check that LUKS PBKDF2 iterations count is non-zero
crypto: strengthen the check for key slots overlapping with LUKS header
crypto: validate that LUKS payload doesn't overlap with header
crypto: enforce that key material doesn't overlap with LUKS header
crypto: enforce that LUKS stripes is always a fixed value
crypto: sanity check that LUKS header strings are NUL-terminated
tests: avoid DOS line endings in PSK file
crypto: check for and report errors setting PSK credentials
scripts: check if .git exists before checking submodule status
seccomp: Get actual errno value from failed seccomp functions
io/channel-watch: Fix socket watch on Windows
io/channel-watch: Drop the unnecessary cast
io/channel-watch: Drop a superfluous '#ifdef WIN32'
util/qemu-sockets: Use g_get_tmp_dir() to get the directory for temporary files
crypto/luks: Support creating LUKS image on Darwin
* tag 'block-pull-request' of https://gitlab.com/stefanha/qemu:
virtio-blk: use BDRV_REQ_REGISTERED_BUF optimization hint
blkio: implement BDRV_REQ_REGISTERED_BUF optimization
stubs: add qemu_ram_block_from_host() and qemu_ram_get_fd()
exec/cpu-common: add qemu_ram_get_fd()
block: add BlockRAMRegistrar
numa: use QLIST_FOREACH_SAFE() for RAM block notifiers
block: return errors from bdrv_register_buf()
block: add BDRV_REQ_REGISTERED_BUF request flag
block: use BdrvRequestFlags type for supported flag fields
block: pass size to bdrv_unregister_buf()
numa: call ->ram_block_removed() in ram_block_notifer_remove()
blkio: add libblkio block driver
coroutine: add flag to re-queue at front of CoQueue
Stefan Hajnoczi [Sun, 30 Oct 2022 19:07:14 +0000 (15:07 -0400)]
Merge tag 'qga-pull-2022-10-26' of https://github.com/kostyanf14/qemu into staging
qga-pull-2022-10-26
# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCgAdFiEEwsLBCepDxjwUI+uE711egWG6hOcFAmNZcDMACgkQ711egWG6
# hOdRPQ/8D5T9GpkC+Ar9GLlx/DZFZx0+FxdolV06TcuSXTz3SXYtVKJUSaLjW3SF
# ZRMbBYxqO1NKXWfe8kPKzSR7lJY9fBdn75AUbN88iYW7xu2x+A8LNskVOanu8vbo
# eqcjO1x3sStzdZHfA4uFUsF7J9A2zqflmZxTMWm1UDt1HTklAJLPkb/E6DnVc3Rp
# WhI89JWvAZBOZoVBq7MyfBPEyf3KibHHMWENVY7vGmmXaw9EJQYpXNEMTeBP1VI3
# tTLxrr8WoGr5w2K4a3Kku2ixD+IOPPWXbZXmSjAGgOwiVSkORwMbUVHnN/A11O3O
# b8XOGZ5LkFjgORTsm9ePxXJvcHlsxWZIb80ZnZA9oGF/33S7RbW1Kcl6OAVGdm0S
# ZzysqPLPJStxuUvesqmFfGkvZ29EHNlWjrIfXQz5mnlDsnOi/0Bus0vNjc/kBNPF
# KfHW6MNw4A2gmVcrNg2f2rlYveHa6e+4XsS5xJJ74WDMHWxMZRaFXyO+qU6p7lEJ
# I5pLD9oEM856y2CpQJpoJnZ6ddLEag652x08WpCFMI76x5XLN8D0lszlgJghR7M7
# 84TKSF6HBvrMS+jQc3xZZlK7ELSluKtUFnSTkNZg9WiuQC/FS+7XRcC6lek1qBAL
# WVsKL5UwM/Eyq+O/B8R7w6TEWFxnEFmjddq1rD6kH8LYx1Lmt4M=
# =h41f
# -----END PGP SIGNATURE-----
# gpg: Signature made Wed 26 Oct 2022 13:36:51 EDT
# gpg: using RSA key C2C2C109EA43C63C1423EB84EF5D5E8161BA84E7
# gpg: Good signature from "Kostiantyn Kostiuk (Upstream PR sign) <[email protected]>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: C2C2 C109 EA43 C63C 1423 EB84 EF5D 5E81 61BA 84E7
* tag 'qga-pull-2022-10-26' of https://github.com/kostyanf14/qemu:
qga: add channel path to error messages
qga: Add HW address getting for FreeBSD
qga: Move HW address getting to a separate function
qga: Add support for user password setting in FreeBSD
qga: Add shutdown/halt/reboot support for FreeBSD
qga: Add UFS freeze/thaw support for FreeBSD
qga: Move Linux-specific FS freeze/thaw code to a separate file
qga: Add initial FreeBSD support
Laurent Vivier [Fri, 21 Oct 2022 09:09:19 +0000 (11:09 +0200)]
qemu-sockets: update socket_uri() and socket_parse() to be consistent
To be consistent with socket_uri(), add 'tcp:' prefix for inet type in
socket_parse(), by default socket_parse() use tcp when no prefix is
provided (format is host:port).
In socket_uri(), use 'vsock:' prefix for vsock type rather than 'tcp:'
because it makes a vsock address look like an inet address with CID
misinterpreted as host.
Goes back to commit 9aca82ba31 "migration: Create socket-address parameter"
Laurent Vivier [Fri, 21 Oct 2022 09:09:11 +0000 (11:09 +0200)]
qapi: net: add stream and dgram netdevs
Copied from socket netdev file and modified to use SocketAddress
to be able to introduce new features like unix socket.
"udp" and "mcast" are squashed into dgram netdev, multicast is detected
according to the IP address type.
"listen" and "connect" modes are managed by stream netdev. An optional
parameter "server" defines the mode (off by default)
The two new types need to be parsed the modern way with -netdev, because
with the traditional way, the "type" field of netdev structure collides with
the "type" field of SocketAddress and prevents the correct evaluation of the
command line option. Moreover the traditional way doesn't allow to use
the same type (SocketAddress) several times with the -netdev option
(needed to specify "local" and "remote" addresses).
The previous commit paved the way for parsing the modern way, but
omitted one detail: how to pick modern vs. traditional, in
netdev_is_modern().
We want to pick based on the value of parameter "type". But how to
extract it from the option argument?
Parsing the option argument, either the modern or the traditional way,
extracts it for us, but only if parsing succeeds.
If parsing fails, there is no good option. No matter which parser we
pick, it'll be the wrong one for some arguments, and the error
reporting will be confusing.
Fortunately, the traditional parser accepts *anything* when called in
a certain way. This maximizes our chance to extract the value of
"type", and in turn minimizes the risk of confusing error reporting.
Laurent Vivier [Fri, 21 Oct 2022 09:09:09 +0000 (11:09 +0200)]
qapi: net: introduce a way to bypass qemu_opts_parse_noisily()
As qemu_opts_parse_noisily() flattens the QAPI structures ("type" field
of Netdev structure can collides with "type" field of SocketAddress),
we introduce a way to bypass qemu_opts_parse_noisily() and use directly
visit_type_Netdev() to parse the backend parameters.
More details from Markus:
qemu_init() passes the argument of -netdev, -nic, and -net to
net_client_parse().
net_client_parse() parses with qemu_opts_parse_noisily(), passing
QemuOptsList qemu_netdev_opts for -netdev, qemu_nic_opts for -nic, and
qemu_net_opts for -net. Their desc[] are all empty, which means any
keys are accepted. The result of the parse (a QemuOpts) is stored in
the QemuOptsList.
Note that QemuOpts is flat by design. In some places, we layer non-flat
on top using dotted keys convention, but not here.
net_init_clients() iterates over the stored QemuOpts, and passes them to
net_init_netdev(), net_param_nic(), or net_init_client(), respectively.
These functions pass the QemuOpts to net_client_init(). They also do
other things with the QemuOpts, which we can ignore here.
net_client_init() uses the opts visitor to convert the (flat) QemOpts to
a (non-flat) QAPI object Netdev. Netdev is also the argument of QMP
command netdev_add.
The opts visitor was an early attempt to support QAPI in
(QemuOpts-based) CLI. It restricts QAPI types to a certain shape; see
commit eb7ee2cbeb "qapi: introduce OptsVisitor".
A more modern way to support QAPI is qobject_input_visitor_new_str().
It uses keyval_parse() instead of QemuOpts for KEY=VALUE,... syntax, and
it also supports JSON syntax. The former isn't quite as expressive as
JSON, but it's a lot closer than QemuOpts + opts visitor.
This commit paves the way to use of the modern way instead.
net: improve error message for missing netdev backend
The current message when using '-net user...' with SLIRP disabled at
compile time is:
qemu-system-x86_64: -net user: Parameter 'type' expects a net backend type (maybe it is not compiled into this binary)
An observation is that we're using the 'netdev->type' field here which
is an enum value, produced after QAPI has converted from its string
form.
IOW, at this point in the code, we know that the user's specified
type name was a valid network backend. The only possible scenario that
can make the backend init function be NULL, is if support for that
backend was disabled at build time. Given this, we don't need to caveat
our error message with a 'maybe' hint, we can be totally explicit.
The use of QERR_INVALID_PARAMETER_VALUE doesn't really lend itself to
user friendly error message text. Since this is not used to set a
specific QAPI error class, we can simply stop using this pre-formatted
error text and provide something better.
Thus the new message is:
qemu-system-x86_64: -net user: network backend 'user' is not compiled into this binary
The case of passing 'hubport' for -net is also given a message reminding
people they should have used -netdev/-nic instead, as this backend type
is only valid for the modern syntax.
Si-Wei Liu [Sat, 8 Oct 2022 07:58:58 +0000 (00:58 -0700)]
vhost-vdpa: allow passing opened vhostfd to vhost-vdpa
Similar to other vhost backends, vhostfd can be passed to vhost-vdpa
backend as another parameter to instantiate vhost-vdpa net client.
This would benefit the use case where only open file descriptors, as
opposed to raw vhost-vdpa device paths, are accessible from the QEMU
process.
Laurent Vivier [Thu, 20 Oct 2022 09:58:45 +0000 (11:58 +0200)]
virtio-net: fix bottom-half packet TX on asynchronous completion
When virtio-net is used with the socket netdev backend, the backend
can be busy and not able to collect new packets.
In this case, net_socket_receive() returns 0 and registers a poll function
to detect when the socket is ready again.
In virtio_net_tx_bh(), virtio_net_flush_tx() forwards the 0, the virtio
notifications are disabled and the function is not re-scheduled, waiting
for the backend to be ready.
When the socket netdev backend is again able to send packets, the poll
function re-starts to flush remaining packets. This is done by
calling virtio_net_tx_complete(). It re-enables notifications and calls
again virtio_net_flush_tx().
But it seems if virtio_net_flush_tx() reaches the tx_burst value all
the queue is not flushed and no new notification is sent to re-schedule
virtio_net_tx_bh(). Nothing re-start to flush the queue and remaining
packets are stuck in the queue.
To fix that, detect in virtio_net_tx_complete() if virtio_net_flush_tx()
has been stopped by tx_burst and if yes re-schedule the bottom half
function virtio_net_tx_bh() to flush the remaining packets.
This is what is done in virtio_net_tx_bh() when the virtio_net_flush_tx()
is synchronous, and completly by-passed when the operation needs to be
asynchronous.
block/block-backend: blk_set_enable_write_cache is IO_CODE
blk_set_enable_write_cache() is defined as GLOBAL_STATE_CODE
but can be invoked from iothreads when handling scsi requests.
This triggers an assertion failure:
0x00007fd6c3515ce1 in raise () from /lib/x86_64-linux-gnu/libc.so.6
0x00007fd6c34ff537 in abort () from /lib/x86_64-linux-gnu/libc.so.6
0x00007fd6c34ff40f in ?? () from /lib/x86_64-linux-gnu/libc.so.6
0x00007fd6c350e662 in __assert_fail () from /lib/x86_64-linux-gnu/libc.so.6
0x000056149e2cea03 in blk_set_enable_write_cache (wce=true, blk=0x5614a01c27f0)
at ../src/block/block-backend.c:1949
0x000056149e2d0a67 in blk_set_enable_write_cache (blk=0x5614a01c27f0,
wce=<optimized out>) at ../src/block/block-backend.c:1951
0x000056149dfe9c59 in scsi_disk_apply_mode_select (p=0x7fd6b400c00e "\004",
page=<optimized out>, s=<optimized out>) at ../src/hw/scsi/scsi-disk.c:1520
mode_select_pages (change=true, len=18, p=0x7fd6b400c00e "\004", r=0x7fd6b4001ff0)
at ../src/hw/scsi/scsi-disk.c:1570
scsi_disk_emulate_mode_select (inbuf=<optimized out>, r=0x7fd6b4001ff0) at
../src/hw/scsi/scsi-disk.c:1640
scsi_disk_emulate_write_data (req=0x7fd6b4001ff0) at ../src/hw/scsi/scsi-disk.c:1934
0x000056149e18ff16 in virtio_scsi_handle_cmd_req_submit (req=<optimized out>,
req=<optimized out>, s=0x5614a12f16b0) at ../src/hw/scsi/virtio-scsi.c:719
virtio_scsi_handle_cmd_vq (vq=0x7fd6bab92140, s=0x5614a12f16b0) at
../src/hw/scsi/virtio-scsi.c:761
virtio_scsi_handle_cmd (vq=<optimized out>, vdev=<optimized out>) at
../src/hw/scsi/virtio-scsi.c:775
virtio_scsi_handle_cmd (vdev=0x5614a12f16b0, vq=0x7fd6bab92140) at
../src/hw/scsi/virtio-scsi.c:765
0x000056149e1a8aa6 in virtio_queue_notify_vq (vq=0x7fd6bab92140) at
../src/hw/virtio/virtio.c:2365
0x000056149e3ccea5 in aio_dispatch_handler (ctx=ctx@entry=0x5614a01babe0,
node=<optimized out>) at ../src/util/aio-posix.c:369
0x000056149e3cd868 in aio_dispatch_ready_handlers (ready_list=0x7fd6c09b2680,
ctx=0x5614a01babe0) at ../src/util/aio-posix.c:399
aio_poll (ctx=0x5614a01babe0, blocking=blocking@entry=true) at
../src/util/aio-posix.c:713
0x000056149e2a7796 in iothread_run (opaque=opaque@entry=0x56149ffde500) at
../src/iothread.c:67
0x000056149e3d0859 in qemu_thread_start (args=0x7fd6c09b26f0) at
../src/util/qemu-thread-posix.c:504
0x00007fd6c36b9ea7 in start_thread () from /lib/x86_64-linux-gnu/libpthread.so.0
0x00007fd6c35d9aef in clone () from /lib/x86_64-linux-gnu/libc.so.6
Changing GLOBAL_STATE_CODE in IO_CODE is allowed, since GSC callers are
allowed to call IO_CODE.
Bin Meng [Thu, 8 Sep 2022 13:28:15 +0000 (21:28 +0800)]
block/nfs: Fix 32-bit Windows build
libnfs.h declares nfs_fstat() as the following for win32:
int nfs_fstat(struct nfs_context *nfs, struct nfsfh *nfsfh,
struct __stat64 *st);
The 'st' parameter should be of type 'struct __stat64'. The
codes happen to build successfully for 64-bit Windows, but it
does not build for 32-bit Windows.
block: use the new _change_ API instead of _can_set_ and _set_
Replace all direct usage of ->can_set_aio_ctx and ->set_aio_ctx,
and call bdrv_child_try_change_aio_context() in
bdrv_try_set_aio_context(), the main function called through
the whole block layer.
From this point onwards, ->can_set_aio_ctx and ->set_aio_ctx
won't be used anymore.
block-backend: implement .change_aio_ctx in child_root
blk_root_change_aio_ctx() is very similar to blk_root_can_set_aio_ctx(),
but implements a new transaction so that if all check pass, the new
transaction's .commit will take care of changing the BlockBackend
AioContext. blk_root_set_aio_ctx_commit() is the same as
blk_root_set_aio_ctx().
Note: bdrv_child_try_change_aio_context() is not called by
anyone at this point.
projects / qemu.git / log