qemu.git
2 months agoMerge remote-tracking branch 'remotes/stefanberger/tags/pull-tpm-2021-01-25-1' into... 246881253 246989393
Peter Maydell [Tue, 26 Jan 2021 09:51:02 +0000 (09:51 +0000)] 
Merge remote-tracking branch 'remotes/stefanberger/tags/pull-tpm-2021-01-25-1' into staging

Merge tpm 2021/01/25 v1

# gpg: Signature made Tue 26 Jan 2021 01:58:26 GMT
# gpg:                using RSA key B818B9CADF9089C2D5CEC66B75AD65802A0B4211
# gpg: Good signature from "Stefan Berger <stefanb@linux.vnet.ibm.com>" [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: B818 B9CA DF90 89C2 D5CE  C66B 75AD 6580 2A0B 4211

* remotes/stefanberger/tags/pull-tpm-2021-01-25-1:
  tpm: tpm_spapr: Remove unused tracepoint

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months agotpm: tpm_spapr: Remove unused tracepoint
Roman Bolshakov [Mon, 7 Dec 2020 06:43:52 +0000 (09:43 +0300)] 
tpm: tpm_spapr: Remove unused tracepoint

Linking of qemu-system-ppc64 fails on macOS with dtrace enabled:

  error: probe tpm_spapr_show_buffer doesn't exist
  error: Could not register probes
  ld: error creating dtrace DOF section for architecture x86_64

The failure is explained in 8c8ed03850208e4 ("net/colo: Match is-enabled
probe to tracepoint"). In short, is-enabled probe can't be used without
a matching trace probe. And for this particular case
tpm_util_show_buffer probe should be enabled to print TPM buffer.

Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
2 months agoMerge remote-tracking branch 'remotes/jasowang/tags/net-pull-request' into staging 246525153 246606604
Peter Maydell [Mon, 25 Jan 2021 15:56:13 +0000 (15:56 +0000)] 
Merge remote-tracking branch 'remotes/jasowang/tags/net-pull-request' into staging

# gpg: Signature made Mon 25 Jan 2021 09:05:51 GMT
# gpg:                using RSA key EF04965B398D6211
# gpg: Good signature from "Jason Wang (Jason Wang on RedHat) <jasowang@redhat.com>" [marginal]
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg:          It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 215D 46F4 8246 689E C77F  3562 EF04 965B 398D 6211

* remotes/jasowang/tags/net-pull-request:
  net: checksum: Introduce fine control over checksum type
  net: checksum: Add IP header checksum calculation
  net: checksum: Skip fragmented IP packets
  net: Fix handling of id in netdev_add and netdev_del

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months agoMerge remote-tracking branch 'remotes/gkurz-gitlab/tags/9p-next-pull-request' into... 246449791 246524745
Peter Maydell [Mon, 25 Jan 2021 13:48:38 +0000 (13:48 +0000)] 
Merge remote-tracking branch 'remotes/gkurz-gitlab/tags/9p-next-pull-request' into staging

This fixes a Coverity report and improves the fid reclaim logic.

# gpg: Signature made Mon 25 Jan 2021 09:37:28 GMT
# gpg:                using RSA key B4828BAF943140CEF2A3491071D4D5E5822F73D6
# gpg: Good signature from "Greg Kurz <groug@kaod.org>" [full]
# gpg:                 aka "Gregory Kurz <gregory.kurz@free.fr>" [full]
# gpg:                 aka "[jpeg image of size 3330]" [full]
# Primary key fingerprint: B482 8BAF 9431 40CE F2A3  4910 71D4 D5E5 822F 73D6

* remotes/gkurz-gitlab/tags/9p-next-pull-request:
  9pfs: Convert reclaim list to QSLIST
  9pfs: Improve unreclaim loop
  9pfs: Convert V9fsFidState::fid_list to QSIMPLEQ
  9pfs: Convert V9fsFidState::clunked to bool
  9pfs/proxy: Check return value of proxy_marshal()

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months agoMerge remote-tracking branch 'remotes/philmd-gitlab/tags/sdmmc-20210124' into staging 246387114 246449538
Peter Maydell [Mon, 25 Jan 2021 11:52:00 +0000 (11:52 +0000)] 
Merge remote-tracking branch 'remotes/philmd-gitlab/tags/sdmmc-20210124' into staging

SD/MMC patches

- Various improvements for SD cards in SPI mode (Bin Meng)

# gpg: Signature made Sun 24 Jan 2021 19:16:55 GMT
# gpg:                using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [full]
# Primary key fingerprint: FAAB E75E 1291 7221 DCFD  6BB2 E3E3 2C2C DEAD C0DE

* remotes/philmd-gitlab/tags/sdmmc-20210124:
  hw/sd: sd.h: Cosmetic change of using spaces
  hw/sd: ssi-sd: Use macros for the dummy value and tokens in the transfer
  hw/sd: ssi-sd: Fix the wrong command index for STOP_TRANSMISSION
  hw/sd: ssi-sd: Add a state representing Nac
  hw/sd: ssi-sd: Suffix a data block with CRC16
  util: Add CRC16 (CCITT) calculation routines
  hw/sd: sd: Drop sd_crc16()
  hw/sd: sd: Support CMD59 for SPI mode
  hw/sd: ssi-sd: Fix incorrect card response sequence

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months agonet: checksum: Introduce fine control over checksum type
Bin Meng [Fri, 11 Dec 2020 09:35:12 +0000 (17:35 +0800)] 
net: checksum: Introduce fine control over checksum type

At present net_checksum_calculate() blindly calculates all types of
checksums (IP, TCP, UDP). Some NICs may have a per type setting in
their BDs to control what checksum should be offloaded. To support
such hardware behavior, introduce a 'csum_flag' parameter to the
net_checksum_calculate() API to allow fine control over what type
checksum is calculated.

Existing users of this API are updated accordingly.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
2 months agonet: checksum: Add IP header checksum calculation
Guishan Qin [Fri, 11 Dec 2020 09:35:11 +0000 (17:35 +0800)] 
net: checksum: Add IP header checksum calculation

At present net_checksum_calculate() only calculates TCP/UDP checksum
in an IP packet, but assumes the IP header checksum to be provided
by the software, e.g.: Linux kernel always calculates the IP header
checksum. However this might not always be the case, e.g.: for an IP
checksum offload enabled stack like VxWorks, the IP header checksum
can be zero.

This adds the checksum calculation of the IP header.

Signed-off-by: Guishan Qin <guishan.qin@windriver.com>
Signed-off-by: Yabing Liu <yabing.liu@windriver.com>
Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Jason Wang <jasowang@redhat.com>
2 months agonet: checksum: Skip fragmented IP packets
Markus Carlstedt [Fri, 11 Dec 2020 09:35:10 +0000 (17:35 +0800)] 
net: checksum: Skip fragmented IP packets

To calculate the TCP/UDP checksum we need the whole datagram. Unless
the hardware has some logic to collect all fragments before sending
the whole datagram first, it can only be done by the network stack,
which is normally the case for the NICs we have seen so far.

Skip these fragmented IP packets to avoid checksum corruption.

Signed-off-by: Markus Carlstedt <markus.carlstedt@windriver.com>
Signed-off-by: Bin Meng <bin.meng@windriver.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
2 months agonet: Fix handling of id in netdev_add and netdev_del
Markus Armbruster [Wed, 25 Nov 2020 10:02:20 +0000 (11:02 +0100)] 
net: Fix handling of id in netdev_add and netdev_del

CLI -netdev accumulates in option group "netdev".

Before commit 08712fcb85 "net: Track netdevs in NetClientState rather
than QemuOpt", netdev_add added to the option group, and netdev_del
removed from it, both HMP and QMP.  Thus, every netdev had a
corresponding QemuOpts in this option group.

Commit 08712fcb85 dropped this for QMP netdev_add and both netdev_del.
Now a netdev has a corresponding QemuOpts only when it was created
with CLI or HMP.  Two issues:

* QMP and HMP netdev_del can leave QemuOpts behind, breaking HMP
  netdev_add.  Reproducer:

    $ qemu-system-x86_64 -S -display none -nodefaults -monitor stdio
    QEMU 5.1.92 monitor - type 'help' for more information
    (qemu) netdev_add user,id=net0
    (qemu) info network
    net0: index=0,type=user,net=10.0.2.0,restrict=off
    (qemu) netdev_del net0
    (qemu) info network
    (qemu) netdev_add user,id=net0
    upstream-qemu: Duplicate ID 'net0' for netdev
    Try "help netdev_add" for more information

  Fix by restoring the QemuOpts deletion in qmp_netdev_del(), but with
  a guard, because the QemuOpts need not exist.

* QMP netdev_add loses its "no duplicate ID" check.  Reproducer:

    $ qemu-system-x86_64 -S -display none -qmp stdio
    {"QMP": {"version": {"qemu": {"micro": 92, "minor": 1, "major": 5}, "package": "v5.2.0-rc2-1-g02c1f0142c"}, "capabilities": ["oob"]}}
    {"execute": "qmp_capabilities"}
    {"return": {}}
    {"execute": "netdev_add", "arguments": {"type": "user", "id":"net0"}}
    {"return": {}}
    {"execute": "netdev_add", "arguments": {"type": "user", "id":"net0"}}
    {"return": {}}

  Fix by adding a duplicate ID check to net_client_init1() to replace
  the lost one.  The check is redundant for callers where QemuOpts
  still checks, i.e. for CLI and HMP.

Reported-by: Andrew Melnichenko <andrew@daynix.com>
Fixes: 08712fcb851034228b61f75bd922863a984a4f60
Cc: qemu-stable@nongnu.org
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
2 months agoMerge remote-tracking branch 'remotes/rth-gitlab/tags/pull-tcg-20210124' into staging 246127551 246347599
Peter Maydell [Sun, 24 Jan 2021 19:36:45 +0000 (19:36 +0000)] 
Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-tcg-20210124' into staging

Fix tcg constant temp overflow.
Fix running during atomic single-step.
Partial support for apple silicon.
Cleanups for accel/tcg.

# gpg: Signature made Sun 24 Jan 2021 18:08:57 GMT
# gpg:                using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg:                issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [full]
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A  05C0 64DF 38E8 AF7E 215F

* remotes/rth-gitlab/tags/pull-tcg-20210124:
  tcg: Restart code generation when we run out of temps
  tcg: Toggle page execution for Apple Silicon
  accel/tcg: Restrict cpu_io_recompile() from other accelerators
  accel/tcg: Declare missing cpu_loop_exit*() stubs
  accel/tcg: Restrict tb_gen_code() from other accelerators
  accel/tcg: Move tb_flush_jmp_cache() to cputlb.c
  accel/tcg: Make cpu_gen_init() static
  tcg: Optimize inline dup_const for MO_64
  qemu/compiler: Split out qemu_build_not_reached_always
  tcg: update the cpu running flag in cpu_exec_step_atomic

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months agohw/sd: sd.h: Cosmetic change of using spaces
Bin Meng [Sat, 23 Jan 2021 10:40:05 +0000 (18:40 +0800)] 
hw/sd: sd.h: Cosmetic change of using spaces

QEMU coding convention prefers spaces over tabs.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210123104016.17485-15-bmeng.cn@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
2 months agohw/sd: ssi-sd: Use macros for the dummy value and tokens in the transfer
Bin Meng [Sat, 23 Jan 2021 10:40:02 +0000 (18:40 +0800)] 
hw/sd: ssi-sd: Use macros for the dummy value and tokens in the transfer

At present the codes use hardcoded numbers (0xff/0xfe) for the dummy
value and block start token. Replace them with macros.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210123104016.17485-12-bmeng.cn@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
2 months agohw/sd: ssi-sd: Fix the wrong command index for STOP_TRANSMISSION
Bin Meng [Sat, 23 Jan 2021 10:40:00 +0000 (18:40 +0800)] 
hw/sd: ssi-sd: Fix the wrong command index for STOP_TRANSMISSION

This fixes the wrong command index for STOP_TRANSMISSION, the
required command to interrupt the multiple block read command,
in the old codes. It should be CMD12 (0x4c), not CMD13 (0x4d).

Fixes: 775616c3ae8c ("Partial SD card SPI mode support")
Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210123104016.17485-10-bmeng.cn@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
2 months agohw/sd: ssi-sd: Add a state representing Nac
Bin Meng [Sat, 23 Jan 2021 10:39:59 +0000 (18:39 +0800)] 
hw/sd: ssi-sd: Add a state representing Nac

Per the "Physical Layer Specification Version 8.00" chapter 7.5.2,
"Data Read", there is a minimum 8 clock cycles (Nac) after the card
response and before data block shows up on the data out line. This
applies to both single and multiple block read operations.

Current implementation of single block read already satisfies the
timing requirement as in the RESPONSE state after all responses are
transferred the state remains unchanged. In the next 8 clock cycles
it jumps to DATA_START state if data is ready.

However we need an explicit state when expanding our support to
multiple block read in the future. Let's add a new state PREP_DATA
explicitly in the ssi-sd state machine to represent Nac.

Note we don't change the single block read state machine to let it
jump from RESPONSE state to DATA_START state as that effectively
generates a 16 clock cycles Nac, which might not be safe. As the
spec says the maximum Nac shall be calculated from several fields
encoded in the CSD register, we don't want to bother updating CSD
to ensure our Nac is within range to complicate things.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210123104016.17485-9-bmeng.cn@gmail.com>
[PMD: Change VMState version id 4 -> 5]
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
2 months agohw/sd: ssi-sd: Suffix a data block with CRC16
Bin Meng [Sat, 23 Jan 2021 10:39:58 +0000 (18:39 +0800)] 
hw/sd: ssi-sd: Suffix a data block with CRC16

Per the SD spec, a valid data block is suffixed with a 16-bit CRC
generated by the standard CCITT polynomial x16+x12+x5+1. This part
is currently missing in the ssi-sd state machine. Without it, all
data block transfer fails in guest software because the expected
CRC16 is missing on the data out line.

Fixes: 775616c3ae8c ("Partial SD card SPI mode support")
Signed-off-by: Bin Meng <bin.meng@windriver.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210123104016.17485-8-bmeng.cn@gmail.com>
[PMD: Change VMState version id 3 -> 4,
      check s->mode validity in post_load()]
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
2 months agoutil: Add CRC16 (CCITT) calculation routines
Bin Meng [Sat, 23 Jan 2021 10:39:57 +0000 (18:39 +0800)] 
util: Add CRC16 (CCITT) calculation routines

Import CRC16 calculation routines from Linux kernel v5.10:

  include/linux/crc-ccitt.h
  lib/crc-ccitt.c

to QEMU:

  include/qemu/crc-ccitt.h
  util/crc-ccitt.c

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20210123104016.17485-7-bmeng.cn@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
[PMD: Restrict compilation to system emulation]
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
2 months agohw/sd: sd: Drop sd_crc16()
Bin Meng [Sat, 23 Jan 2021 10:39:56 +0000 (18:39 +0800)] 
hw/sd: sd: Drop sd_crc16()

commit f6fb1f9b319f ("sdcard: Correct CRC16 offset in sd_function_switch()")
changed the 16-bit CRC to be stored at offset 64. In fact, this CRC
calculation is completely wrong. From the original codes, it wants
to calculate the CRC16 of the first 64 bytes of sd->data[], however
passing 64 as the `width` to sd_crc16() actually counts 256 bytes
starting from the `message` for the CRC16 calculation, which is not
what we want.

Besides that, it seems existing sd_crc16() algorithm does not match
the SD spec, which says CRC16 is the CCITT one but the calculation
does not produce expected result. It turns out the CRC16 was never
transferred outside the sd core, as in sd_read_byte() we see:

    if (sd->data_offset >= 64)
        sd->state = sd_transfer_state;

Given above reasons, let's drop it.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Pragnesh Patel <pragnesh.patel@sifive.com>
Reviewed-by: Pragnesh Patel <pragnesh.patel@sifive.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210123104016.17485-6-bmeng.cn@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
2 months agotcg: Restart code generation when we run out of temps
Richard Henderson [Sat, 23 Jan 2021 22:11:17 +0000 (12:11 -1000)] 
tcg: Restart code generation when we run out of temps

Some large translation blocks can generate so many unique
constants that we run out of temps to hold them.  In this
case, longjmp back to the start of code generation and
restart with a smaller translation block.

Buglink: https://bugs.launchpad.net/bugs/1912065
Tested-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months agohw/sd: sd: Support CMD59 for SPI mode
Bin Meng [Sat, 23 Jan 2021 10:39:55 +0000 (18:39 +0800)] 
hw/sd: sd: Support CMD59 for SPI mode

After the card is put into SPI mode, CRC check for all commands
including CMD0 will be done according to CMD59 setting. But this
command is currently unimplemented. Simply allow the decoding of
CMD59, but the CRC remains unchecked.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Pragnesh Patel <pragnesh.patel@sifive.com>
Reviewed-by: Pragnesh Patel <pragnesh.patel@sifive.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210123104016.17485-5-bmeng.cn@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
2 months agohw/sd: ssi-sd: Fix incorrect card response sequence
Bin Meng [Sat, 23 Jan 2021 10:39:54 +0000 (18:39 +0800)] 
hw/sd: ssi-sd: Fix incorrect card response sequence

Per the "Physical Layer Specification Version 8.00" chapter 7.5.1,
"Command/Response", there is a minimum 8 clock cycles (Ncr) before
the card response shows up on the data out line. However current
implementation jumps directly to the sending response state after
all 6 bytes command is received, which is a spec violation.

Add a new state PREP_RESP in the ssi-sd state machine to handle it.

Fixes: 775616c3ae8c ("Partial SD card SPI mode support")
Signed-off-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Pragnesh Patel <pragnesh.patel@sifive.com>
Reviewed-by: Pragnesh Patel <pragnesh.patel@sifive.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210123104016.17485-4-bmeng.cn@gmail.com>
[PMD: Change VMState version id 2 -> 3]
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
2 months agoMerge remote-tracking branch 'remotes/bonzini-gitlab/tags/for-upstream' into staging 245943454 246101152
Peter Maydell [Sat, 23 Jan 2021 22:34:21 +0000 (22:34 +0000)] 
Merge remote-tracking branch 'remotes/bonzini-gitlab/tags/for-upstream' into staging

* Make MinGW respect --bindir (Joshua)
* Switch MinGW to a "deep" installation (Joshua + NSIS fixes by myself)
* Fix compilation errors/warnings (Qixin, Philippe)
* QemuOpts cleanups (myself)
* Consistency improvements for -action (myself)
* remove deprecated "change vnc TARGET" functionality (myself)
* meson cleanups (Marc-André, Philippe, myself)
* IDE out-of-bounds access (Prasad)
* LA57 fix for -cpu max (Weijiang)

# gpg: Signature made Sat 23 Jan 2021 20:55:59 GMT
# gpg:                using RSA key F13338574B662389866C7682BFFBD25F78C7AE83
# gpg:                issuer "pbonzini@redhat.com"
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>" [full]
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>" [full]
# Primary key fingerprint: 46F5 9FBD 57D6 12E7 BFD4  E2F7 7E15 100C CD36 69B1
#      Subkey fingerprint: F133 3857 4B66 2389 866C  7682 BFFB D25F 78C7 AE83

* remotes/bonzini-gitlab/tags/for-upstream: (30 commits)
  qemu-option: warn for short-form boolean options
  qemu-option: move help handling to get_opt_name_value
  qemu-option: clean up id vs. list->merge_lists
  vnc: support "-vnc help"
  qmp: remove deprecated "change" command
  hmp: remove "change vnc TARGET" command
  acceptance: switch to QMP change-vnc-password command
  meson.build: Detect bzip2 program
  meson.build: Declare global edk2_targets / install_edk2_blobs variables
  meson: Add a section header for library dependencies
  meson: Display crypto-related information altogether
  meson: Display block layer information altogether
  meson: Display accelerators and selected targets altogether
  meson: Summarize compilation-related information altogether
  meson: Summarize overall features altogether
  meson: Display host binaries information altogether
  meson: Summarize information related to directories first
  meson: convert wixl detection to Meson
  nsis: adjust for new MinGW paths
  meson: Declare have_virtfs_proxy_helper in main meson.build
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months agotcg: Toggle page execution for Apple Silicon
Roman Bolshakov [Wed, 13 Jan 2021 03:28:07 +0000 (06:28 +0300)] 
tcg: Toggle page execution for Apple Silicon

Pages can't be both write and executable at the same time on Apple
Silicon. macOS provides public API to switch write protection [1] for
JIT applications, like TCG.

1. https://developer.apple.com/documentation/apple_silicon/porting_just-in-time_compilers_to_apple_silicon

Tested-by: Alexander Graf <agraf@csgraf.de>
Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
Message-Id: <20210113032806.18220-1-r.bolshakov@yadro.com>
[rth: Inline the qemu_thread_jit_* functions;
 drop the MAP_JIT change for a follow-on patch.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months agoaccel/tcg: Restrict cpu_io_recompile() from other accelerators
Philippe Mathieu-Daudé [Sun, 17 Jan 2021 16:48:12 +0000 (17:48 +0100)] 
accel/tcg: Restrict cpu_io_recompile() from other accelerators

As cpu_io_recompile() is only called within TCG accelerator
in cputlb.c, declare it locally.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210117164813.4101761-6-f4bug@amsat.org>
[rth: Adjust vs changed tb_flush_jmp_cache patch.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months agoaccel/tcg: Declare missing cpu_loop_exit*() stubs
Philippe Mathieu-Daudé [Sun, 17 Jan 2021 16:48:11 +0000 (17:48 +0100)] 
accel/tcg: Declare missing cpu_loop_exit*() stubs

cpu_loop_exit*() functions are declared in accel/tcg/cpu-exec-common.c,
and are not available when TCG accelerator is not built. Add stubs so
linking without TCG succeed.

Problematic files:

- hw/semihosting/console.c in qemu_semihosting_console_inc()
- hw/ppc/spapr_hcall.c in h_confer()
- hw/s390x/ipl.c in s390_ipl_reset_request()
- hw/misc/mips_itu.c

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210117164813.4101761-5-f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months agoaccel/tcg: Restrict tb_gen_code() from other accelerators
Philippe Mathieu-Daudé [Thu, 21 Jan 2021 06:15:06 +0000 (20:15 -1000)] 
accel/tcg: Restrict tb_gen_code() from other accelerators

tb_gen_code() is only called within TCG accelerator, declare it locally.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210117164813.4101761-4-f4bug@amsat.org>
[rth: Adjust vs changed tb_flush_jmp_cache patch.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months agoaccel/tcg: Move tb_flush_jmp_cache() to cputlb.c
Richard Henderson [Thu, 21 Jan 2021 05:53:20 +0000 (19:53 -1000)] 
accel/tcg: Move tb_flush_jmp_cache() to cputlb.c

Move and make the function static, as the only users
are here in cputlb.c.

Suggested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months agoaccel/tcg: Make cpu_gen_init() static
Philippe Mathieu-Daudé [Sun, 17 Jan 2021 16:48:08 +0000 (17:48 +0100)] 
accel/tcg: Make cpu_gen_init() static

cpu_gen_init() is TCG specific, only used in tcg/translate-all.c.
No need to export it to other accelerators, declare it statically.

Reviewed-by: Claudio Fontana <cfontana@suse.de>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210117164813.4101761-2-f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months agoqemu-option: warn for short-form boolean options
Paolo Bonzini [Mon, 9 Nov 2020 09:13:39 +0000 (04:13 -0500)] 
qemu-option: warn for short-form boolean options

Options such as "server" or "nowait", that are commonly found in -chardev,
are sugar for "server=on" and "wait=off".  This is quite surprising and
also does not have any notion of typing attached.  It is even possible to
do "-device e1000,noid" and get a device with "id=off".

Deprecate it and print a warning when it is encountered.  In general,
this short form for boolean options only seems to be in wide use for
-chardev and -spice.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agoqemu-option: move help handling to get_opt_name_value
Paolo Bonzini [Tue, 3 Nov 2020 13:48:11 +0000 (08:48 -0500)] 
qemu-option: move help handling to get_opt_name_value

Right now, help options are parsed normally and then checked
specially in opt_validate, but only if coming from
qemu_opts_parse_noisily.  has_help_option does the check on its own.

opt_validate() has two callers: qemu_opt_set(), which passes null and is
therefore unaffected, and opts_do_parse(), which is affected.

opts_do_parse() is called by qemu_opts_do_parse(), which passes null and
is therefore unaffected, and opts_parse().

opts_parse() is called by qemu_opts_parse() and qemu_opts_set_defaults(),
which pass null and are therefore unaffected, and
qemu_opts_parse_noisily().

Move the check from opt_validate to the parsing workhorse of QemuOpts,
get_opt_name_value.  This will come in handy in the next patch, which
will raise a warning for "-object memory-backend-ram,share" ("flag" option
with no =on/=off part) but not for "-object memory-backend-ram,help".

As a result:

- opts_parse and opts_do_parse do not return an error anymore
  when help is requested; qemu_opts_parse_noisily does not have
  to work around that anymore.

- various crazy ways to request help are not recognized anymore:
  - "help=..."
  - "nohelp" (sugar for "help=off")
  - "?=..."
  - "no?" (sugar for "?=off")

- "help" would be recognized as help request even if there is a (foolishly
  named) parameter "help".  No such parameters exist, though.

Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agoqemu-option: clean up id vs. list->merge_lists
Paolo Bonzini [Mon, 9 Nov 2020 09:58:24 +0000 (04:58 -0500)] 
qemu-option: clean up id vs. list->merge_lists

Looking at all merge-lists QemuOptsList, here is how they access their
QemuOpts:

reopen_opts in qemu-io-cmds.c ("qemu-img reopen -o")
qemu_opts_find(&reopen_opts, NULL)

empty_opts in qemu-io.c ("qemu-io open -o")
qemu_opts_find(&empty_opts, NULL)

qemu_rtc_opts ("-rtc")
qemu_find_opts_singleton("rtc")

qemu_machine_opts ("-M")
qemu_find_opts_singleton("machine")

qemu_action_opts ("-name")
qemu_opts_foreach->process_runstate_actions

qemu_boot_opts ("-boot")
in hw/nvram/fw_cfg.c and hw/s390x/ipl.c:
  QTAILQ_FIRST(&qemu_find_opts("bootopts")->head)
in softmmu/vl.c:
  qemu_opts_find(qemu_find_opts("boot-opts"), NULL)

qemu_name_opts ("-name")
qemu_opts_foreach->parse_name
parse_name does not use id

qemu_mem_opts ("-m")
qemu_find_opts_singleton("memory")

qemu_icount_opts ("-icount")
qemu_opts_foreach->do_configure_icount
do_configure_icount->icount_configure
icount_configure does not use id

qemu_smp_opts ("-smp")
qemu_opts_find(qemu_find_opts("smp-opts"), NULL)

qemu_spice_opts ("-spice")
QTAILQ_FIRST(&qemu_spice_opts.head)

i.e. they don't need an id.  Sometimes its presence is ignored
(e.g. when using qemu_opts_foreach), sometimes all the options
with the id are skipped, sometimes only the first option on the
command line is considered.  -boot does two different things
depending on who's looking at the options.

With this patch we just forbid id on merge-lists QemuOptsLists; if the
command line still works, it has the same semantics as before.

qemu_opts_create's fail_if_exists parameter is now unnecessary:

- it is unused if id is NULL

- opts_parse only passes false if reached from qemu_opts_set_defaults,
in which case this patch enforces that id must be NULL

- other callers that can pass a non-NULL id always set it to true

Assert that it is true in the only case where "fail_if_exists" matters,
i.e. "id && !lists->merge_lists".  This means that if an id is present,
duplicates are always forbidden, which was already the status quo.

Discounting the case that aborts as it's not user-controlled (it's
"just" a matter of inspecting qemu_opts_create callers), the paths
through qemu_opts_create can be summarized as:

- merge_lists = true: singleton opts with NULL id; non-NULL id fails

- merge_lists = false: always return new opts; non-NULL id fails if dup

Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agovnc: support "-vnc help"
Paolo Bonzini [Wed, 20 Jan 2021 14:42:35 +0000 (15:42 +0100)] 
vnc: support "-vnc help"

Use qemu_opts_parse_noisily now that HMP does not call
vnc_parse anymore.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20210120144235.345983-4-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agoqmp: remove deprecated "change" command
Paolo Bonzini [Wed, 20 Jan 2021 14:42:34 +0000 (15:42 +0100)] 
qmp: remove deprecated "change" command

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210120144235.345983-3-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agohmp: remove "change vnc TARGET" command
Paolo Bonzini [Wed, 20 Jan 2021 14:42:33 +0000 (15:42 +0100)] 
hmp: remove "change vnc TARGET" command

The HMP command \"change vnc TARGET\" is messy:

- it takes an ugly shortcut to determine if the option has an "id",
with incorrect results if "id=" is not preceded by an unescaped
comma.

- it deletes the existing QemuOpts and does not try to rollback
if the parsing fails (which is not causing problems, but only due to
how VNC options are parsed)

- because it uses the same parsing function as "-vnc", it forces
the latter to not support "-vnc help".

On top of this, it uses a deprecated QMP command, thus getting in
the way of removing the QMP command.  Since the usecase for the
command is not clear, just remove it and send "change vnc password"
directly to the QMP "change-vnc-password" command.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20210120144235.345983-2-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agoacceptance: switch to QMP change-vnc-password command
Paolo Bonzini [Thu, 21 Jan 2021 13:55:12 +0000 (08:55 -0500)] 
acceptance: switch to QMP change-vnc-password command

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson.build: Detect bzip2 program
Philippe Mathieu-Daudé [Thu, 14 Jan 2021 17:45:09 +0000 (18:45 +0100)] 
meson.build: Detect bzip2 program

The --enable-bzip2/--disable-bzip2 configure arguments are
somehow misleading, they check for the bzip2 library, not
the bzip2 program.

We need the bzip2 program to install the EDK2 firmware blobs
(see commit 623ef637a2e "configure: Check bzip2 is available").

Check if the bzip2 program in the global meson.build to avoid
the configuration to succeed, but a later when trying to install
the firmware blobs:

    ../pc-bios/meson.build:5:2: ERROR: Program 'bzip2' not found

Reported-by: John Snow <jsnow@redhat.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Fixes: c8d5450bba3 ("configure: move install_blobs from configure to meson")
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210114174509.2944817-3-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson.build: Declare global edk2_targets / install_edk2_blobs variables
Philippe Mathieu-Daudé [Thu, 14 Jan 2021 17:45:08 +0000 (18:45 +0100)] 
meson.build: Declare global edk2_targets / install_edk2_blobs variables

Globally declare in the main meson.build:
- the list of EDK2 targets,
- whether the EDK2 blobs have to be installed.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210114174509.2944817-2-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: Add a section header for library dependencies
Philippe Mathieu-Daudé [Thu, 21 Jan 2021 09:56:16 +0000 (10:56 +0100)] 
meson: Add a section header for library dependencies

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210121095616.1471869-9-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: Display crypto-related information altogether
Philippe Mathieu-Daudé [Thu, 21 Jan 2021 09:56:15 +0000 (10:56 +0100)] 
meson: Display crypto-related information altogether

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210121095616.1471869-8-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: Display block layer information altogether
Philippe Mathieu-Daudé [Thu, 21 Jan 2021 09:56:14 +0000 (10:56 +0100)] 
meson: Display block layer information altogether

Display block layer information altogether,
when it is relevant.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210121095616.1471869-7-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: Display accelerators and selected targets altogether
Philippe Mathieu-Daudé [Thu, 21 Jan 2021 09:56:13 +0000 (10:56 +0100)] 
meson: Display accelerators and selected targets altogether

Display accelerators and selected targets altogether,
avoid to display unuseful information when not relevant.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210121095616.1471869-6-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: Summarize compilation-related information altogether
Philippe Mathieu-Daudé [Thu, 21 Jan 2021 09:56:12 +0000 (10:56 +0100)] 
meson: Summarize compilation-related information altogether

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210121095616.1471869-5-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: Summarize overall features altogether
Philippe Mathieu-Daudé [Thu, 21 Jan 2021 09:56:11 +0000 (10:56 +0100)] 
meson: Summarize overall features altogether

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210121095616.1471869-4-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: Display host binaries information altogether
Philippe Mathieu-Daudé [Thu, 21 Jan 2021 09:56:10 +0000 (10:56 +0100)] 
meson: Display host binaries information altogether

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210121095616.1471869-3-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: Summarize information related to directories first
Philippe Mathieu-Daudé [Thu, 21 Jan 2021 09:56:09 +0000 (10:56 +0100)] 
meson: Summarize information related to directories first

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210121095616.1471869-2-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: convert wixl detection to Meson
Paolo Bonzini [Thu, 21 Jan 2021 10:49:04 +0000 (11:49 +0100)] 
meson: convert wixl detection to Meson

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agonsis: adjust for new MinGW paths
Paolo Bonzini [Thu, 21 Jan 2021 12:17:13 +0000 (07:17 -0500)] 
nsis: adjust for new MinGW paths

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agomeson: Declare have_virtfs_proxy_helper in main meson.build
Philippe Mathieu-Daudé [Wed, 20 Jan 2021 15:15:39 +0000 (16:15 +0100)] 
meson: Declare have_virtfs_proxy_helper in main meson.build

have_virtfs_proxy_helper is used from docs/meson.build, and can be
not declared when including it before fsdev/meson.build. This fixes:

  ../docs/meson.build:54:2: ERROR: Unknown variable "have_virtfs_proxy_helper".

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210120151539.1166252-1-philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agoMerge remote-tracking branch 'remotes/kraxel/tags/usb-20210122-pull-request' into... 245873398 245916096
Peter Maydell [Sat, 23 Jan 2021 14:40:45 +0000 (14:40 +0000)] 
Merge remote-tracking branch 'remotes/kraxel/tags/usb-20210122-pull-request' into staging

usb: cleanups and fixes.
usb: add pcap support.

# gpg: Signature made Fri 22 Jan 2021 17:48:35 GMT
# gpg:                using RSA key A0328CFFB93A17A79901FE7D4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>" [full]
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>" [full]
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>" [full]
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/usb-20210122-pull-request:
  usb-host: map LIBUSB_SPEED_SUPER_PLUS to USB_SPEED_SUPER
  usb: add pcap support.
  hw/usb/dev-uas: Report command additional adb length as unsupported
  scsi/utils: Add INVALID_PARAM_VALUE sense code definition
  hw/usb/hcd-xhci: Fix extraneous format-truncation error on 32-bit hosts
  hw/usb: Convert to qdev_realize()
  hw/usb: Fix bad printf format specifiers
  hw/usb/host-libusb.c: fix build with kernel < 5.0

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months agobuild-sys: add libvhost-user missing dependencies
Marc-André Lureau [Thu, 14 Jan 2021 12:56:03 +0000 (16:56 +0400)] 
build-sys: add libvhost-user missing dependencies

This help fixing static compilation issues.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210114125605.1227742-4-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agobuild-sys: set global arguments for cflags/ldflags
Marc-André Lureau [Thu, 14 Jan 2021 12:56:02 +0000 (16:56 +0400)] 
build-sys: set global arguments for cflags/ldflags

As we want subprojects to share those arguments

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20210114125605.1227742-3-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agobuild-sys: remove unused LIBS
Marc-André Lureau [Thu, 14 Jan 2021 12:56:01 +0000 (16:56 +0400)] 
build-sys: remove unused LIBS

Since meson switch and commit feabc71dfa58 ("configure: do not include
dependency flags in QEMU_CFLAGS and LIBS"), LIBS is unused.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20210114125605.1227742-2-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agoide: atapi: check logical block address and read size (CVE-2020-29443)
Prasad J Pandit [Mon, 18 Jan 2021 11:51:30 +0000 (17:21 +0530)] 
ide: atapi: check logical block address and read size (CVE-2020-29443)

While processing ATAPI cmd_read/cmd_read_cd commands,
Logical Block Address (LBA) maybe invalid OR closer to the last block,
leading to an OOB access issues. Add range check to avoid it.

Fixes: CVE-2020-29443
Reported-by: Wenxiang Qian <leonwxqian@gmail.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org>
Message-Id: <20210118115130.457044-1-ppandit@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agosoftmmu/physmem: Silence GCC 10 maybe-uninitialized error
Philippe Mathieu-Daudé [Sun, 17 Jan 2021 17:04:11 +0000 (18:04 +0100)] 
softmmu/physmem: Silence GCC 10 maybe-uninitialized error

When building with GCC 10.2 configured with --extra-cflags=-Os, we get:

  softmmu/physmem.c: In function 'address_space_translate_for_iotlb':
  softmmu/physmem.c:643:26: error: 'notifier' may be used uninitialized in this function [-Werror=maybe-uninitialized]
    643 |         notifier->active = true;
        |                          ^
  softmmu/physmem.c:608:23: note: 'notifier' was declared here
    608 |     TCGIOMMUNotifier *notifier;
        |                       ^~~~~~~~

Initialize 'notifier' to silence the warning.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210117170411.4106949-1-f4bug@amsat.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agotcg: Optimize inline dup_const for MO_64
Richard Henderson [Sun, 16 Feb 2020 21:43:10 +0000 (13:43 -0800)] 
tcg: Optimize inline dup_const for MO_64

Avoid the out-of-line function call for immediate MO_64.
In addition, diagnose all invalid constants at compile-time.

Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months agoqemu/compiler: Split out qemu_build_not_reached_always
Richard Henderson [Thu, 14 Jan 2021 20:04:04 +0000 (10:04 -1000)] 
qemu/compiler: Split out qemu_build_not_reached_always

Provide a symbol that can always be used to signal an error,
regardless of optimization.  Usage of this should be protected
by e.g. __builtin_constant_p, which guards for optimization.

Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months agotcg: update the cpu running flag in cpu_exec_step_atomic
Douglas Crosher [Tue, 22 Sep 2020 07:42:41 +0000 (17:42 +1000)] 
tcg: update the cpu running flag in cpu_exec_step_atomic

The cpu_exec_step_atomic() function is called with the cpu->running
clear and proceeds to run target code without setting this flag. If
this target code generates an exception then handle_cpu_signal() will
unnecessarily abort.  For example if atomic code generates a memory
protection fault.

This patch at least sets and clears this running flag, and adds some
assertions to help detect other cases.

Signed-off-by: Douglas Crosher <dtc-ubuntu@scieneer.com>
Message-Id: <a272c656-f7c5-019d-1cc0-499b8f80f2fc@scieneer.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 months ago9pfs: Convert reclaim list to QSLIST
Greg Kurz [Fri, 22 Jan 2021 14:35:14 +0000 (15:35 +0100)] 
9pfs: Convert reclaim list to QSLIST

Use QSLIST instead of open-coding for a slightly improved readability.

No behavioral change.

Reviewed-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Message-Id: <20210122143514.215780-1-groug@kaod.org>
Signed-off-by: Greg Kurz <groug@kaod.org>
2 months ago9pfs: Improve unreclaim loop
Greg Kurz [Thu, 21 Jan 2021 18:15:10 +0000 (19:15 +0100)] 
9pfs: Improve unreclaim loop

If a fid was actually re-opened by v9fs_reopen_fid(), we re-traverse the
fid list from the head in case some other request created a fid that
needs to be marked unreclaimable as well (i.e. the client opened a new
handle on the path that is being unlinked). This is suboptimal since
most if not all fids that require it have likely been taken care of
already.

This is mostly the result of new fids being added to the head of the
list. Since the list is now a QSIMPLEQ, add new fids at the end instead
to avoid the need to rewind. Take a reference on the fid to ensure it
doesn't go away during v9fs_reopen_fid() and that it can be safely
passed to QSIMPLEQ_NEXT() afterwards. Since the associated put_fid()
can also yield, same is done with the next fid. So the logic here is
to get a reference on a fid and only put it back during the next
iteration after we could get a reference on the next fid.

Reviewed-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Message-Id: <20210121181510.1459390-1-groug@kaod.org>
Signed-off-by: Greg Kurz <groug@kaod.org>
2 months agousb-host: map LIBUSB_SPEED_SUPER_PLUS to USB_SPEED_SUPER
Gerd Hoffmann [Thu, 21 Jan 2021 15:08:32 +0000 (16:08 +0100)] 
usb-host: map LIBUSB_SPEED_SUPER_PLUS to USB_SPEED_SUPER

Handle host superspeedplus (usb 3.1+) devices like superspeed (usb 3.0)
devices.  That is enough to get them handled properly by xhci.  They show
up as superspeed devices inside the guest, but should be able to actually
run at higher speeds.

Reported-by: Angel Pagan <Angel.Pagan@stratus.com>
Tested-by: Angel Pagan <Angel.Pagan@stratus.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20210121150832.3564097-1-kraxel@redhat.com>

2 months agousb: add pcap support.
Gerd Hoffmann [Tue, 19 Jan 2021 19:44:51 +0000 (20:44 +0100)] 
usb: add pcap support.

Log all traffic of a specific usb device to a pcap file for later
inspection.  File format is compatible with linux usb monitor.

Usage:
  qemu -device usb-${somedevice},pcap=file.pcap
  wireshark file.pcap

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210119194452.2148048-1-kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
2 months agohw/usb/dev-uas: Report command additional adb length as unsupported
Philippe Mathieu-Daudé [Wed, 20 Jan 2021 15:35:22 +0000 (16:35 +0100)] 
hw/usb/dev-uas: Report command additional adb length as unsupported

We are not ready to handle additional CDB data.

If a guest sends a packet with such additional data,
report the command parameter as not supported.

Specify a size (of 1 byte) for the add_cdb member we
are not using, to fix the following warning:

  usb/dev-uas.c:157:31: error: field 'status' with variable sized type 'uas_iu' not at the end of a struct or class is a GNU extension [-Werror,-Wgnu-variable-sized-type-not-at-end]
      uas_iu                    status;
                                ^

Reported-by: Ed Maste <emaste@FreeBSD.org>
Reported-by: Daniele Buono <dbuono@linux.vnet.ibm.com>
Reported-by: Han Han <hhan@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210120153522.1173897-4-philmd@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
2 months agoscsi/utils: Add INVALID_PARAM_VALUE sense code definition
Philippe Mathieu-Daudé [Wed, 20 Jan 2021 15:35:21 +0000 (16:35 +0100)] 
scsi/utils: Add INVALID_PARAM_VALUE sense code definition

Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210120153522.1173897-3-philmd@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
2 months agohw/usb/hcd-xhci: Fix extraneous format-truncation error on 32-bit hosts
Philippe Mathieu-Daudé [Mon, 18 Jan 2021 18:11:15 +0000 (19:11 +0100)] 
hw/usb/hcd-xhci: Fix extraneous format-truncation error on 32-bit hosts

For some reason the assert() added in commit ccb799313a5
("hw/usb: avoid format truncation warning when formatting
port name") does not fix when building with GCC 10.

KISS and expand the buffer by 4 bytes to silent the following
error when using GCC 10.2.1 on Fedora 33:

  hw/usb/hcd-xhci.c: In function 'usb_xhci_realize':
  hw/usb/hcd-xhci.c:3309:54: error: '%d' directive output may be truncated writing between 1 and 8 bytes into a region of size 5 [-Werror=format-truncation=]
   3309 |             snprintf(port->name, sizeof(port->name), "usb2 port #%d", i+1);
        |                                                      ^~~~~~~~~~~~~~~
  hw/usb/hcd-xhci.c:3309:54: note: directive argument in the range [1, 89478486]
  In file included from /usr/include/stdio.h:866,
                   from include/qemu/osdep.h:85,
                   from hw/usb/hcd-xhci.c:22:
  /usr/include/bits/stdio2.h:70:10: note: '__builtin___snprintf_chk' output between 13 and 20 bytes into a destination of size 16
     70 |   return __builtin___snprintf_chk (__s, __n, __USE_FORTIFY_LEVEL - 1,
        |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     71 |        __bos (__s), __fmt, __va_arg_pack ());
        |        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  hw/usb/hcd-xhci.c:3323:54: error: '%d' directive output may be truncated writing between 1 and 8 bytes into a region of size 5 [-Werror=format-truncation=]
   3323 |             snprintf(port->name, sizeof(port->name), "usb3 port #%d", i+1);
        |                                                      ^~~~~~~~~~~~~~~
  hw/usb/hcd-xhci.c:3323:54: note: directive argument in the range [1, 89478486]
  In file included from /usr/include/stdio.h:866,
                   from include/qemu/osdep.h:85,
                   from hw/usb/hcd-xhci.c:22:
  /usr/include/bits/stdio2.h:70:10: note: '__builtin___snprintf_chk' output between 13 and 20 bytes into a destination of size 16
     70 |   return __builtin___snprintf_chk (__s, __n, __USE_FORTIFY_LEVEL - 1,
        |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     71 |        __bos (__s), __fmt, __va_arg_pack ());
        |        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  cc1: all warnings being treated as errors

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210118181115.313742-1-philmd@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
2 months agohw/usb: Convert to qdev_realize()
Markus Armbruster [Tue, 19 Jan 2021 12:01:51 +0000 (13:01 +0100)] 
hw/usb: Convert to qdev_realize()

Device code shouldn't mess with QOM property "realized" since we have
proper interfaces (merge commit 6675a653).  Commit 8ddab8dd3d
"usb/hcd-xhci: Split pci wrapper for xhci base model" and commit
f00ff136ee "usb: hcd-xhci-sysbus: Attach xhci to sysbus device"
reintroduced two instances.  Clean them up.  Note that s->xhci is
a (bus-less) TYPE_XHCI device.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210119120151.53757-1-armbru@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
2 months agohw/usb: Fix bad printf format specifiers
Alex Chen [Thu, 19 Nov 2020 02:57:51 +0000 (02:57 +0000)] 
hw/usb: Fix bad printf format specifiers

We should use printf format specifier "%u" instead of "%d" for
argument of type "unsigned int".

Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Alex Chen <alex.chen@huawei.com>
Message-id: 20201119025751.45750-1-alex.chen@huawei.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
2 months agohw/usb/host-libusb.c: fix build with kernel < 5.0
Fabrice Fontaine [Sun, 13 Dec 2020 21:30:16 +0000 (22:30 +0100)] 
hw/usb/host-libusb.c: fix build with kernel < 5.0

USBDEVFS_GET_SPEED is used since version 5.2.0 and
https://gitlab.com/qemu-project/qemu/-/commit/202d69a715a4b1824dcd7ec1683d027ed2bae6d3
resulting in the following build failure with kernel < 5.0:

../hw/usb/host-libusb.c: In function 'usb_host_open':
../hw/usb/host-libusb.c:953:32: error: 'USBDEVFS_GET_SPEED' undeclared (first use in this function); did you mean 'USBDEVFS_GETDRIVER'?
         int rc = ioctl(hostfd, USBDEVFS_GET_SPEED, NULL);
                                ^~~~~~~~~~~~~~~~~~
                                USBDEVFS_GETDRIVER

A tentative was made to fix this build failure with
https://gitlab.com/qemu-project/qemu/-/commit/4969e697c15ac536d5c0700381d5d026ef7f0588

However, the assumption that distros with old kernels also have old
libusb is just wrong so also add a check for defined(USBDEVFS_GET_SPEED)

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Message-id: 20201213213016.457350-1-fontaine.fabrice@gmail.com

[ kraxel: codestyle whitespace fixup ]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
2 months agoMerge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-6.0-pull-request... 245405650 245484926
Peter Maydell [Fri, 22 Jan 2021 10:35:55 +0000 (10:35 +0000)] 
Merge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-6.0-pull-request' into staging

linux-user pull request 20210119-v3

Remove obsolete F_SHLCK and F_EXLCK translation
Update sockopt
Add F_ADD_SEALS and F_GET_SEALS

# gpg: Signature made Thu 21 Jan 2021 12:33:04 GMT
# gpg:                using RSA key CD2F75DDC8E3A4DC2E4F5173F30C38BD3F2FBE3C
# gpg:                issuer "laurent@vivier.eu"
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>" [full]
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>" [full]
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>" [full]
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier2/tags/linux-user-for-6.0-pull-request:
  linux-user: Remove obsolete F_SHLCK and F_EXLCK translation
  linux-user: Add IPv6 options to do_print_sockopt()
  linux-user: add missing IPv6 get/setsockopt option
  linux-user: add missing UDP get/setsockopt option
  linux-user: Support F_ADD_SEALS and F_GET_SEALS fcntls

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months agoMerge remote-tracking branch 'remotes/cohuck-gitlab/tags/s390x-20210121' into staging 245122167 245403746
Peter Maydell [Thu, 21 Jan 2021 19:09:45 +0000 (19:09 +0000)] 
Merge remote-tracking branch 'remotes/cohuck-gitlab/tags/s390x-20210121' into staging

s390x updates:
- headers update to Linux 5.11-rc2
- fix tcg emulation for some instructions that are generated by
  clang Linux kernel builds
- vfio-ccw: wire up the device unplug notification mechanism
- fix a gcc 11 warning

# gpg: Signature made Thu 21 Jan 2021 12:08:12 GMT
# gpg:                using RSA key C3D0D66DC3624FF6A8C018CEDECF6B93C6F02FAF
# gpg:                issuer "cohuck@redhat.com"
# gpg: Good signature from "Cornelia Huck <conny@cornelia-huck.de>" [unknown]
# gpg:                 aka "Cornelia Huck <huckc@linux.vnet.ibm.com>" [full]
# gpg:                 aka "Cornelia Huck <cornelia.huck@de.ibm.com>" [full]
# gpg:                 aka "Cornelia Huck <cohuck@kernel.org>" [unknown]
# gpg:                 aka "Cornelia Huck <cohuck@redhat.com>" [unknown]
# Primary key fingerprint: C3D0 D66D C362 4FF6 A8C0  18CE DECF 6B93 C6F0 2FAF

* remotes/cohuck-gitlab/tags/s390x-20210121:
  s390x: Use strpadcpy for copying vm name
  vfio-ccw: Connect the device request notifier
  Update linux headers to 5.11-rc2
  update-linux-headers: Include const.h
  s390x/tcg: Ignore register content if b1/b2 is zero when handling EXECUTE
  tests/tcg/s390x: Fix EXRL tests
  s390x/tcg: Don't ignore content in r0 when not specified via "b" or "x"
  s390x/tcg: Fix RISBHG
  s390x/tcg: Fix ALGSI

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months ago9pfs: Convert V9fsFidState::fid_list to QSIMPLEQ
Greg Kurz [Mon, 18 Jan 2021 14:22:59 +0000 (15:22 +0100)] 
9pfs: Convert V9fsFidState::fid_list to QSIMPLEQ

The fid_list is currently open-coded. This doesn't seem to serve any
purpose that cannot be met with QEMU's generic lists. Let's go for a
QSIMPLEQ : this will allow to add new fids at the end of the list and
to improve the logic in v9fs_mark_fids_unreclaim().

Reviewed-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Message-Id: <20210118142300.801516-3-groug@kaod.org>
Signed-off-by: Greg Kurz <groug@kaod.org>
2 months ago9pfs: Convert V9fsFidState::clunked to bool
Greg Kurz [Mon, 18 Jan 2021 14:22:58 +0000 (15:22 +0100)] 
9pfs: Convert V9fsFidState::clunked to bool

This can only be 0 or 1.

Reviewed-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Message-Id: <20210118142300.801516-2-groug@kaod.org>
Signed-off-by: Greg Kurz <groug@kaod.org>
2 months ago9pfs/proxy: Check return value of proxy_marshal()
Greg Kurz [Mon, 11 Jan 2021 09:49:56 +0000 (10:49 +0100)] 
9pfs/proxy: Check return value of proxy_marshal()

This should always successfully write exactly two 32-bit integers.
Make it clear with an assert(), like v9fs_receive_status() and
v9fs_receive_response() already do when unmarshalling the same
header.

Fixes: Coverity CID 1438968
Reviewed-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Message-Id: <161035859647.1221144.4691749806675653934.stgit@bahia.lan>
Signed-off-by: Greg Kurz <groug@kaod.org>
2 months agolinux-user: Remove obsolete F_SHLCK and F_EXLCK translation
Michael Forney [Thu, 14 Jan 2021 22:36:02 +0000 (14:36 -0800)] 
linux-user: Remove obsolete F_SHLCK and F_EXLCK translation

These lock types are unsupported by Linux since v2.2[0][1] and
always return EINVAL (except on SPARC up until v2.6, which just
warned).

musl libc does not define these constants, so just remove them from
the translation cases.

[0] https://github.com/mpe/linux-fullhistory/blob/v2.2.0/fs/locks.c#L322-L324
[1] https://github.com/mpe/linux-fullhistory/blob/v2.2.0/fs/locks.c#L429-L445

Signed-off-by: Michael Forney <mforney@mforney.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210114223602.9004-1-mforney@mforney.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 months agolinux-user: Add IPv6 options to do_print_sockopt()
Shu-Chun Weng [Fri, 18 Dec 2020 19:32:13 +0000 (11:32 -0800)] 
linux-user: Add IPv6 options to do_print_sockopt()

Signed-off-by: Shu-Chun Weng <scw@google.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20201218193213.3566856-5-scw@google.com>
[lv: Add missing <linux/in6.h> include for IPV6_ADDR_PREFERENCES]
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 months agoutil/cacheflush: Fix error generated by clang
Gan Qixin [Fri, 15 Jan 2021 07:56:56 +0000 (15:56 +0800)] 
util/cacheflush: Fix error generated by clang

When compiling qemu-fuzz-i386 on aarch64 host, clang reported the following
error:

../util/cacheflush.c:38:44: error: value size does not match register size
specified by the constraint and modifier [-Werror,-Wasm-operand-widths]
    asm volatile("mrs\t%0, ctr_el0" : "=r"(save_ctr_el0));
                                           ^
../util/cacheflush.c:38:24: note: use constraint modifier "w"
    asm volatile("mrs\t%0, ctr_el0" : "=r"(save_ctr_el0));
                       ^~
                       %w0

Modify the type of save_ctr_el0 to uint64_t to fix it.

Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Gan Qixin <ganqixin@huawei.com>
Message-Id: <20210115075656.717957-1-ganqixin@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
2 months agobuild-system: clean up TCG/TCI configury
Paolo Bonzini [Mon, 14 Dec 2020 11:01:45 +0000 (12:01 +0100)] 
build-system: clean up TCG/TCI configury

Make CONFIG_TCG_INTERPRETER a Meson option, and enable TCI (though with
a warning) if the host CPU is unsupported, making it more similar to
other --enable-* options.

Remove TCG-specific include paths from !CONFIG_TCG builds.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agox86/cpu: Use max host physical address if -cpu max option is applied
Yang Weijiang [Wed, 13 Jan 2021 09:04:30 +0000 (17:04 +0800)] 
x86/cpu: Use max host physical address if -cpu max option is applied

QEMU option -cpu max(max_features) means "Enables all features supported by
the accelerator in the current host", this looks true for all the features
except guest max physical address width, so add this patch to enable it.

Signed-off-by: Yang Weijiang <weijiang.yang@intel.com>
Message-Id: <20210113090430.26394-1-weijiang.yang@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agoconfigure: MinGW respect --bindir argument
Joshua Watt [Tue, 12 Jan 2021 21:02:39 +0000 (15:02 -0600)] 
configure: MinGW respect --bindir argument

There are two cases that need to be accounted for when compiling QEMU
for MinGW32:
 1) A standalone distribution, where QEMU is self contained and
    extracted by the user, such as a user would download from the QEMU
    website. In this case, all the QEMU executable files should be
    rooted in $prefix to ensure they can be easily found by the user
 2) QEMU integrated into a distribution image/sysroot/SDK and
    distributed with other programs. In this case, the provided
    arguments for bindir/datadir/etc. should be respected as they for a
    Linux build.

Restructures the MinGW path configuration so that all of the paths
except bindir use the same rules as when building for other platforms.
This satisfies #2 and #1 since these files do not need to be directly in
$prefix anyway.

The handling for --bindir is changed so that it defaults to $prefix on
MinGW (maintaining the compatibility with #1), but if the user specifies
a specific path when configuring it can also satisfy #2.

Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Message-Id: <20210112210239.28836-1-JPEWhacker@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agorunstate: cleanup reboot and panic actions
Paolo Bonzini [Wed, 20 Jan 2021 13:30:27 +0000 (14:30 +0100)] 
runstate: cleanup reboot and panic actions

The possible choices for panic, reset and watchdog actions are inconsistent.

"-action panic=poweroff" should be renamed to "-action panic=shutdown"
on the command line.  This is because "-action panic=poweroff" and
"-action watchdog=poweroff" have slightly different semantics, the first
does an unorderly exit while the second goes through qemu_cleanup().  With
this change, -no-shutdown would not have to change "-action panic=pause"
"pause", just like it does not have to change the reset action.

"-action reboot=none" should be renamed to "-action reboot=reset".
This should be self explanatory, since for example "-action panic=none"
lets the guest proceed without taking any action.

Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 months agoMerge remote-tracking branch 'remotes/ericb/tags/pull-nbd-2021-01-20' into staging 244852969 244880378
Peter Maydell [Thu, 21 Jan 2021 10:44:27 +0000 (10:44 +0000)] 
Merge remote-tracking branch 'remotes/ericb/tags/pull-nbd-2021-01-20' into staging

nbd patches for 2021-01-20

- minor resource leak fixes in qemu-nbd
- ensure proper aio context when nbd server uses iothreads
- iotest refactorings in preparation for rewriting ./check to be more
flexible, and preparing for more nbd server reconnect features

# gpg: Signature made Thu 21 Jan 2021 02:28:19 GMT
# gpg:                using RSA key 71C2CC22B1C4602927D2F3AAA7A16B4A2527436A
# gpg: Good signature from "Eric Blake <eblake@redhat.com>" [full]
# gpg:                 aka "Eric Blake (Free Software Programmer) <ebb9@byu.net>" [full]
# gpg:                 aka "[jpeg image of size 6874]" [full]
# Primary key fingerprint: 71C2 CC22 B1C4 6029 27D2  F3AA A7A1 6B4A 2527 436A

* remotes/ericb/tags/pull-nbd-2021-01-20:
  iotests.py: qemu_io(): reuse qemu_tool_pipe_and_status()
  iotests.py: fix qemu_tool_pipe_and_status()
  iotests/264: fix style
  iotests: define group in each iotest
  iotests/294: add shebang line
  iotests: make tests executable
  iotests: fix some whitespaces in test output files
  iotests/303: use dot slash for qcow2.py running
  iotests/277: use dot slash for nbd-fault-injector.py running
  nbd/server: Quiesce coroutines on context switch
  block: Honor blk_set_aio_context() context requirements
  qemu-nbd: Fix a memleak in nbd_client_thread()
  qemu-nbd: Fix a memleak in qemu_nbd_client_list()

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 months agos390x: Use strpadcpy for copying vm name
Miroslav Rezanina [Thu, 14 Jan 2021 07:07:36 +0000 (08:07 +0100)] 
s390x: Use strpadcpy for copying vm name

Using strncpy with length equal to the size of target array, GCC 11
reports following warning:

  warning: '__builtin_strncpy' specified bound 256 equals destination size [-Wstringop-truncation]

We can prevent this warning by using strpadcpy that copies string
up to specified length, zeroes target array after copied string
and does not raise warning when length is equal to target array
size (and ending '\0' is discarded).

Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <6f86915755219cf6a671788075da4809b57f7d7b.1610607906.git.mrezanin@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
2 months agovfio-ccw: Connect the device request notifier
Eric Farman [Mon, 4 Jan 2021 20:20:57 +0000 (21:20 +0100)] 
vfio-ccw: Connect the device request notifier

Now that the vfio-ccw code has a notifier interface to request that
a device be unplugged, let's wire that together.

Signed-off-by: Eric Farman <farman@linux.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <20210104202057.48048-4-farman@linux.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
2 months agoUpdate linux headers to 5.11-rc2
Eric Farman [Mon, 4 Jan 2021 20:20:56 +0000 (21:20 +0100)] 
Update linux headers to 5.11-rc2

Signed-off-by: Eric Farman <farman@linux.ibm.com>
Message-Id: <20210104202057.48048-3-farman@linux.ibm.com>
[CH: dropped qatomic->atomic changes in pvrdma_ring.h]
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
2 months agoupdate-linux-headers: Include const.h
Eric Farman [Mon, 4 Jan 2021 20:20:55 +0000 (21:20 +0100)] 
update-linux-headers: Include const.h

Kernel commit a85cbe6159ff ("uapi: move constants from
<linux/kernel.h> to <linux/const.h>") breaks our script
because of the unrecognized include. Let's add that to
our processing.

Signed-off-by: Eric Farman <farman@linux.ibm.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210104202057.48048-2-farman@linux.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
2 months agos390x/tcg: Ignore register content if b1/b2 is zero when handling EXECUTE
David Hildenbrand [Mon, 11 Jan 2021 16:38:45 +0000 (17:38 +0100)] 
s390x/tcg: Ignore register content if b1/b2 is zero when handling EXECUTE

In our EXECUTE fast path, we have to ignore the content of r0, if
specified by b1 or b2.

Fixes: d376f123c7de ("target/s390x: Re-implement a few EXECUTE target insns directly")
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20210111163845.18148-6-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
2 months agotests/tcg/s390x: Fix EXRL tests
David Hildenbrand [Mon, 11 Jan 2021 16:38:44 +0000 (17:38 +0100)] 
tests/tcg/s390x: Fix EXRL tests

The current EXRL tests crash on real machines: we must not use r0 as a base
register for trt/trtr, otherwise the content gets ignored. Also, we must
not use r0 for exrl, otherwise it gets ignored.

Let's use the "a" constraint so we get a general purpose register != r0.
For op2, we can simply specify a memory operand directly via "Q" (Memory
reference without index register and with short displacement).

Fixes: ad8c851d2e77 ("target/s390x: add EX support for TRT and TRTR")
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20210111163845.18148-5-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
2 months agos390x/tcg: Don't ignore content in r0 when not specified via "b" or "x"
David Hildenbrand [Mon, 11 Jan 2021 16:38:43 +0000 (17:38 +0100)] 
s390x/tcg: Don't ignore content in r0 when not specified via "b" or "x"

Using get_address() with register identifiers comming from an "r" field
is wrong: if the "r" field designates "r0", we don't read the content
and instead assume 0 - which should only be applied when the register
was specified via "b" or "x".

PoP 5-11 "Operand-Address Generation":
  "A zero in any of the B1, B2, X2, B3, or B4 fields indicates the absence
   of the corresponding address component. For the absent component, a zero
   is used in forming the intermediate sum, regardless of the contents of
   general register 0. A displacement of zero has no special significance."

This BUG became visible for CSPG as generated by LLVM-12 in the upstream
Linux kernel (v5.11-rc2), used while creating the linear mapping in
vmem_map_init(): Trying to store to address 0 results in a Low Address
Protection exception.

Debugging this was more complicated than it could have been: The program
interrupt handler in the kernel will try to crash the kernel: doing so, it
will enable DAT. As the linear mapping is not created yet (asce=0), we run
into an addressing exception while tring to walk non-existant DAT tables,
resulting in a program exception loop.

This allows for booting upstream Linux kernels compiled by clang-12. Most
of these cases seem to be broken forever.

Reported-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Cc: Guenter Roeck <linux@roeck-us.net>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20210111163845.18148-4-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
2 months agos390x/tcg: Fix RISBHG
David Hildenbrand [Mon, 11 Jan 2021 16:38:42 +0000 (17:38 +0100)] 
s390x/tcg: Fix RISBHG

RISBHG is broken and currently hinders clang-11 builds of upstream kernels
from booting: the kernel crashes early, while decompressing the image.

  [...]
   Kernel fault: interruption code 0005 ilc:2
   Kernel random base: 0000000000000000
   PSW : 0000200180000000 0000000000017a1e
         R:0 T:0 IO:0 EX:0 Key:0 M:0 W:0 P:0 AS:0 CC:2 PM:0 RI:0 EA:3
   GPRS: 0000000000000001 0000000c00000000 00000003fffffff4 00000000fffffff0
         0000000000000000 00000000fffffff4 000000000000000c 00000000fffffff0
         00000000fffffffc 0000000000000000 00000000fffffff8 00000000008e25a8
         0000000000000009 0000000000000002 0000000000000008 000000000000bce0

One example of a buggy instruction is:

    17dde:       ec 1e 00 9f 20 5d       risbhg  %r1,%r14,0,159,32

With %r14 = 0x9 and %r1 = 0x7 should result in %r1 = 0x900000007, however,
results in %r1 = 0.

Let's interpret values of i3/i4 as documented in the PoP and make
computation of "mask" only based on i3 and i4 and use "pmask" only at the
very end to make sure wrapping is only applied to the high/low doubleword.

With this patch, I can successfully boot a v5.11-rc2 kernel built with
clang-11, and gcc builds keep on working.

Fixes: 2d6a869833d9 ("target-s390: Implement RISBG")
Reported-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Cc: Guenter Roeck <linux@roeck-us.net>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20210111163845.18148-3-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
2 months agos390x/tcg: Fix ALGSI
David Hildenbrand [Mon, 11 Jan 2021 16:38:41 +0000 (17:38 +0100)] 
s390x/tcg: Fix ALGSI

Looks like something went wrong whiel touching that line. Instead of "r1"
we need a new temporary. Also, we have to pass MO_TEQ, to indicate that
we are working with 64-bit values. Let's revert these changes.

Fixes: ff26d287bddc ("target/s390x: Improve cc computation for ADD LOGICAL")
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20210111163845.18148-2-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
2 months agoiotests.py: qemu_io(): reuse qemu_tool_pipe_and_status()
Vladimir Sementsov-Ogievskiy [Mon, 30 Nov 2020 13:40:20 +0000 (16:40 +0300)] 
iotests.py: qemu_io(): reuse qemu_tool_pipe_and_status()

Just drop code duplication.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201130134024.19212-5-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoiotests.py: fix qemu_tool_pipe_and_status()
Vladimir Sementsov-Ogievskiy [Mon, 30 Nov 2020 13:40:19 +0000 (16:40 +0300)] 
iotests.py: fix qemu_tool_pipe_and_status()

qemu_img_args variable is unrelated here. We should print just args.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201130134024.19212-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoiotests/264: fix style
Vladimir Sementsov-Ogievskiy [Wed, 18 Nov 2020 18:04:28 +0000 (21:04 +0300)] 
iotests/264: fix style

Fix long line, extra import and one mypy complaint about incompatible
int and float.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201118180433.11931-7-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoiotests: define group in each iotest
Vladimir Sementsov-Ogievskiy [Sat, 16 Jan 2021 13:44:19 +0000 (16:44 +0300)] 
iotests: define group in each iotest

We are going to drop group file. Define group in tests as a preparatory
step.

The patch is generated by

    cd tests/qemu-iotests

    grep '^[0-9]\{3\} ' group | while read line; do
        file=$(awk '{print $1}' <<< "$line");
        groups=$(sed -e 's/^... //' <<< "$line");
        awk "NR==2{print \"# group: $groups\"}1" $file > tmp;
        cat tmp > $file;
    done

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210116134424.82867-7-vsementsov@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoiotests/294: add shebang line
Vladimir Sementsov-Ogievskiy [Sat, 16 Jan 2021 13:44:18 +0000 (16:44 +0300)] 
iotests/294: add shebang line

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210116134424.82867-6-vsementsov@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoiotests: make tests executable
Vladimir Sementsov-Ogievskiy [Sat, 16 Jan 2021 13:44:17 +0000 (16:44 +0300)] 
iotests: make tests executable

All other test files are executable. Fix these.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210116134424.82867-5-vsementsov@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoiotests: fix some whitespaces in test output files
Vladimir Sementsov-Ogievskiy [Sat, 16 Jan 2021 13:44:16 +0000 (16:44 +0300)] 
iotests: fix some whitespaces in test output files

We are going to be stricter about comparing test result with .out
files. So, fix some whitespaces now.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210116134424.82867-4-vsementsov@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoiotests/303: use dot slash for qcow2.py running
Vladimir Sementsov-Ogievskiy [Sat, 16 Jan 2021 13:44:15 +0000 (16:44 +0300)] 
iotests/303: use dot slash for qcow2.py running

If you run './check 303', check includes common.config which adjusts
$PATH to include '.' first, and therefore finds qcow2.py on PATH.  But
if you run './303' directly, there is nothing to adjust PATH, and if
'.' is not already on your PATH by other means, the test fails because
the executable is not found.  Adjust how we invoke the helper
executable to avoid needing a PATH search in the first place.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210116134424.82867-3-vsementsov@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoiotests/277: use dot slash for nbd-fault-injector.py running
Vladimir Sementsov-Ogievskiy [Sat, 16 Jan 2021 13:44:14 +0000 (16:44 +0300)] 
iotests/277: use dot slash for nbd-fault-injector.py running

If you run './check 277', check includes common.config which adjusts
$PATH to include '.' first, and therefore finds nbd-fault-injector.py
on PATH.  But if you run './277' directly, there is nothing to adjust
PATH, and if '.' is not already on your PATH by other means, the test
fails because the executable is not found.  Adjust how we invoke the
helper executable to avoid needing a PATH search in the first place.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210116134424.82867-2-vsementsov@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agonbd/server: Quiesce coroutines on context switch
Sergio Lopez [Mon, 14 Dec 2020 17:05:18 +0000 (18:05 +0100)] 
nbd/server: Quiesce coroutines on context switch

When switching between AIO contexts we need to me make sure that both
recv_coroutine and send_coroutine are not scheduled to run. Otherwise,
QEMU may crash while attaching the new context with an error like
this one:

aio_co_schedule: Co-routine was already scheduled in 'aio_co_schedule'

To achieve this we need a local implementation of
'qio_channel_readv_all_eof' named 'nbd_read_eof' (a trick already done
by 'nbd/client.c') that allows us to interrupt the operation and to
know when recv_coroutine is yielding.

With this in place, we delegate detaching the AIO context to the
owning context with a BH ('nbd_aio_detach_bh') scheduled using
'aio_wait_bh_oneshot'. This BH signals that we need to quiesce the
channel by setting 'client->quiescing' to 'true', and either waits for
the coroutine to finish using AIO_WAIT_WHILE or, if it's yielding in
'nbd_read_eof', actively enters the coroutine to interrupt it.

RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=1900326
Signed-off-by: Sergio Lopez <slp@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20201214170519.223781-4-slp@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoblock: Honor blk_set_aio_context() context requirements
Sergio Lopez [Mon, 14 Dec 2020 17:05:16 +0000 (18:05 +0100)] 
block: Honor blk_set_aio_context() context requirements

The documentation for bdrv_set_aio_context_ignore() states this:

 * The caller must own the AioContext lock for the old AioContext of bs, but it
 * must not own the AioContext lock for new_context (unless new_context is the
 * same as the current context of bs).

As blk_set_aio_context() makes use of this function, this rule also
applies to it.

Fix all occurrences where this rule wasn't honored.

Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Sergio Lopez <slp@redhat.com>
Message-Id: <20201214170519.223781-2-slp@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2 months agoqemu-nbd: Fix a memleak in nbd_client_thread()
Alex Chen [Tue, 8 Dec 2020 13:49:44 +0000 (13:49 +0000)] 
qemu-nbd: Fix a memleak in nbd_client_thread()

When the qio_channel_socket_connect_sync() fails
we should goto 'out_socket' label to free the 'sioc' instead of
goto 'out' label.
In addition, there's a lot of redundant code in the successful branch
and the error branch, optimize it.

Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Alex Chen <alex.chen@huawei.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201208134944.27962-1-alex.chen@huawei.com>