guest agent: remove uneeded dependencies

This patch tries to cull any uneeded library dependencies from the guest
agent to improve portability across various distros. We do so by being
as explicit as possible about in-tree dependencies rather than relying
on existing *-obj-y targets, and by manually setting LIBS for the
qemu-ga target to avoid pulling in LIBS_TOOLS libraries discovered by
configure.

Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

guest agent: remove g_strcmp0 usage

g_strcmp0 isn't in all version of glib 2.0, so don't use it to avoid
build breakage on older distros.

Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

memory: crack wide ioport accesses into smaller ones when needed

The memory API supports cracking wide accesses into narrower ones
when needed; but this was no implemented for the pio address space,
causing lsi53c895a's IO BAR to malfunction.

Fix by correctly cracking wide accesses when needed.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

memory: abstract cracking of write access ops into a function

The memory API automatically cracks large reads and writes into smaller
ones when needed.  Factor out this mechanism, which is now duplicated between
memory reads and memory writes, into a function.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

escc: replace DPRINTFs with tracepoints

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

m48t59: avoid structure holes spotted by pahole

Report from pahole on amd64 host:
struct M48t59State {
uint32_t                   type;                 /*     0     4 */

/* XXX 4 bytes hole, try to pack */

qemu_irq                   IRQ;                  /*     8     8 */
uint32_t                   io_base;              /*    16     4 */
uint32_t                   size;                 /*    20     4 */
time_t                     time_offset;          /*    24     8 */
time_t                     stop_time;            /*    32     8 */
struct tm                  alarm;                /*    40    56 */
/* --- cacheline 1 boundary (64 bytes) was 32 bytes ago --- */
struct QEMUTimer *         alrm_timer;           /*    96     8 */
struct QEMUTimer *         wd_timer;             /*   104     8 */
uint8_t                    lock;                 /*   112     1 */

/* XXX 1 byte hole, try to pack */

uint16_t                   addr;                 /*   114     2 */

/* XXX 4 bytes hole, try to pack */

uint8_t *                  buffer;               /*   120     8 */
/* --- cacheline 2 boundary (128 bytes) --- */

/* size: 128, cachelines: 2 */
/* sum members: 119, holes: 3, sum holes: 9 */
}; /* definitions: 1 */

Fix by rearranging the structure to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

escc: avoid structure holes spotted by pahole

Edited report from pahole on amd64 host:
struct ChannelState {
...
ChnType                    type;                 /*    32     4 */

/* XXX 4 bytes hole, try to pack */
...
uint8_t                    rregs[16];            /*    66    16 */

/* XXX 2 bytes hole, try to pack */
...
/* size: 392, cachelines: 7 */
/* sum members: 382, holes: 2, sum holes: 6 */
/* padding: 4 */
/* last cacheline: 8 bytes */
}; /* definitions: 1 */

Fix by rearranging the structure to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

fdc: avoid structure holes spotted by pahole

Edited report from pahole on amd64 host:
struct FDCtrl {
uint8_t                    version;              /*     0     1 */

/* XXX 7 bytes hole, try to pack */

qemu_irq                   irq;                  /*     8     8 */
int                        dma_chann;            /*    16     4 */

/* XXX 4 bytes hole, try to pack */
...
uint8_t                    status2;              /*    42     1 */

/* XXX 5 bytes hole, try to pack */

uint8_t *                  fifo;                 /*    48     8 */
...
uint8_t                    pwrd;                 /*    76     1 */

/* XXX 3 bytes hole, try to pack */

int                        sun4m;                /*    80     4 */
uint8_t                    num_floppies;         /*    84     1 */

/* XXX 3 bytes hole, try to pack */

FDrive                     drives[2];            /*    88    64 */
/* --- cacheline 2 boundary (128 bytes) was 24 bytes ago --- */
int                        reset_sensei;         /*   152     4 */

/* size: 160, cachelines: 3 */
/* sum members: 134, holes: 5, sum holes: 22 */
/* padding: 4 */
/* last cacheline: 32 bytes */
}; /* definitions: 1 */

Fix by rearranging the structure to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

pcnet: avoid structure holes spotted by pahole

Edited report from pahole on amd64 host:
struct PCNetState_st {
...
uint16_t                   bcr[32];              /*   340    64 */

/* XXX 4 bytes hole, try to pack */
...
int                        tx_busy;              /*  4520     4 */

/* XXX 4 bytes hole, try to pack */

qemu_irq                   irq;                  /*  4528     8 */
void                       (*phys_mem_read)(void *, target_phys_addr_t, uint8_t *, int, int); /*  4536     8 */
/* --- cacheline 71 boundary (4544 bytes) --- */
void                       (*phys_mem_write)(void *, target_phys_addr_t, uint8_t *, int, int); /*  4544     8 */
void *                     dma_opaque;           /*  4552     8 */
int                        looptest;             /*  4560     4 */

/* size: 4568, cachelines: 72 */
/* sum members: 4556, holes: 2, sum holes: 8 */
/* padding: 4 */
/* last cacheline: 24 bytes */
}; /* definitions: 2 */

Fix by rearranging the structure to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

esp: avoid structure holes spotted by pahole

Report from pahole on amd64 host:
struct ESPState {
SysBusDevice               busdev;               /*     0  5648 */
/* --- cacheline 88 boundary (5632 bytes) was 16 bytes ago --- */
uint32_t                   it_shift;             /*  5648     4 */

/* XXX 4 bytes hole, try to pack */

qemu_irq                   irq;                  /*  5656     8 */
uint8_t                    rregs[16];            /*  5664    16 */
uint8_t                    wregs[16];            /*  5680    16 */
/* --- cacheline 89 boundary (5696 bytes) --- */
int32_t                    ti_size;              /*  5696     4 */
uint32_t                   ti_rptr;              /*  5700     4 */
uint32_t                   ti_wptr;              /*  5704     4 */
uint8_t                    ti_buf[16];           /*  5708    16 */
uint32_t                   status;               /*  5724     4 */
uint32_t                   dma;                  /*  5728     4 */

/* XXX 4 bytes hole, try to pack */

SCSIBus                    bus;                  /*  5736  2120 */
/* --- cacheline 122 boundary (7808 bytes) was 48 bytes ago --- */
SCSIDevice *               current_dev;          /*  7856     8 */
SCSIRequest *              current_req;          /*  7864     8 */
/* --- cacheline 123 boundary (7872 bytes) --- */
uint8_t                    cmdbuf[16];           /*  7872    16 */
uint32_t                   cmdlen;               /*  7888     4 */
uint32_t                   do_cmd;               /*  7892     4 */
uint32_t                   dma_left;             /*  7896     4 */
uint32_t                   dma_counter;          /*  7900     4 */
uint8_t *                  async_buf;            /*  7904     8 */
uint32_t                   async_len;            /*  7912     4 */

/* XXX 4 bytes hole, try to pack */

ESPDMAMemoryReadWriteFunc  dma_memory_read;      /*  7920     8 */
ESPDMAMemoryReadWriteFunc  dma_memory_write;     /*  7928     8 */
/* --- cacheline 124 boundary (7936 bytes) --- */
void *                     dma_opaque;           /*  7936     8 */
int                        dma_enabled;          /*  7944     4 */

/* XXX 4 bytes hole, try to pack */

void                       (*dma_cb)(ESPState *); /*  7952     8 */

/* size: 7960, cachelines: 125 */
/* sum members: 7944, holes: 4, sum holes: 16 */
/* last cacheline: 24 bytes */
}; /* definitions: 1 */

Fix by rearranging the structure to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

sun4m: avoid structure holes spotted by pahole

Edited report from pahole on amd64 host:
struct sun4c_hwdef {
...
uint8_t                    nvram_machine_id;     /*   112     1 */

/* XXX 1 byte hole, try to pack */
...
/* size: 136, cachelines: 3 */
/* sum members: 135, holes: 1, sum holes: 1 */
/* last cacheline: 8 bytes */
}; /* definitions: 1 */

struct sun4d_hwdef {
...
uint8_t                    nvram_machine_id;     /*   128     1 */

/* XXX 1 byte hole, try to pack */
...
/* size: 152, cachelines: 3 */
/* sum members: 151, holes: 1, sum holes: 1 */
/* last cacheline: 24 bytes */
}; /* definitions: 1 */

struct sun4m_hwdef {
...
uint8_t                    nvram_machine_id;     /*   260     1 */

/* XXX 1 byte hole, try to pack */

uint16_t                   machine_id;           /*   262     2 */
uint32_t                   iommu_version;        /*   264     4 */

/* XXX 4 bytes hole, try to pack */
...
/* size: 288, cachelines: 5 */
/* sum members: 283, holes: 2, sum holes: 5 */
/* last cacheline: 32 bytes */
}; /* definitions: 1 */

Fix by rearranging the structures to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

tcx: avoid structure holes spotted by pahole

Report from pahole on amd64 host:
struct TCXState {
SysBusDevice               busdev;               /*     0  5648 */
/* --- cacheline 88 boundary (5632 bytes) was 16 bytes ago --- */
target_phys_addr_t         addr;                 /*  5648     8 */
DisplayState *             ds;                   /*  5656     8 */
uint8_t *                  vram;                 /*  5664     8 */
uint32_t *                 vram24;               /*  5672     8 */
uint32_t *                 cplane;               /*  5680     8 */
ram_addr_t                 vram_offset;          /*  5688     8 */
/* --- cacheline 89 boundary (5696 bytes) --- */
ram_addr_t                 vram24_offset;        /*  5696     8 */
ram_addr_t                 cplane_offset;        /*  5704     8 */
uint32_t                   vram_size;            /*  5712     4 */
uint16_t                   width;                /*  5716     2 */
uint16_t                   height;               /*  5718     2 */
uint16_t                   depth;                /*  5720     2 */
uint8_t                    r[256];               /*  5722   256 */
/* --- cacheline 93 boundary (5952 bytes) was 26 bytes ago --- */
uint8_t                    g[256];               /*  5978   256 */
/* --- cacheline 97 boundary (6208 bytes) was 26 bytes ago --- */
uint8_t                    b[256];               /*  6234   256 */

/* XXX 2 bytes hole, try to pack */

/* --- cacheline 101 boundary (6464 bytes) was 28 bytes ago --- */
uint32_t                   palette[256];         /*  6492  1024 */
/* --- cacheline 117 boundary (7488 bytes) was 28 bytes ago --- */
uint8_t                    dac_index;            /*  7516     1 */
uint8_t                    dac_state;            /*  7517     1 */

/* size: 7520, cachelines: 118 */
/* sum members: 7516, holes: 1, sum holes: 2 */
/* padding: 2 */
/* last cacheline: 32 bytes */
}; /* definitions: 1 */

Fix by rearranging the structure to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

sun4m_iommu: avoid structure holes spotted by pahole

Report from pahole on amd64 host:
struct IOMMUState {
SysBusDevice               busdev;               /*     0  5648 */
/* --- cacheline 88 boundary (5632 bytes) was 16 bytes ago --- */
uint32_t                   regs[4096];           /*  5648 16384 */
/* --- cacheline 344 boundary (22016 bytes) was 16 bytes ago --- */
target_phys_addr_t         iostart;              /* 22032     8 */
uint32_t                   version;              /* 22040     4 */

/* XXX 4 bytes hole, try to pack */

qemu_irq                   irq;                  /* 22048     8 */

/* size: 22056, cachelines: 345 */
/* sum members: 22052, holes: 1, sum holes: 4 */
/* last cacheline: 40 bytes */
}; /* definitions: 1 */

Fix by rearranging the structure to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

slavio_intctl: avoid structure holes spotted by pahole

Report from pahole on amd64 host:
struct SLAVIO_INTCTLState {
SysBusDevice               busdev;               /*     0  5648 */
/* --- cacheline 88 boundary (5632 bytes) was 16 bytes ago --- */
uint32_t                   intregm_pending;      /*  5648     4 */
uint32_t                   intregm_disabled;     /*  5652     4 */
uint32_t                   target_cpu;           /*  5656     4 */

/* XXX 4 bytes hole, try to pack */

qemu_irq                   cpu_irqs[16][16];     /*  5664  2048 */
/* --- cacheline 120 boundary (7680 bytes) was 32 bytes ago --- */
SLAVIO_CPUINTCTLState      slaves[16];           /*  7712   384 */
/* --- cacheline 126 boundary (8064 bytes) was 32 bytes ago --- */

/* size: 8096, cachelines: 127 */
/* sum members: 8092, holes: 1, sum holes: 4 */
/* last cacheline: 32 bytes */
}; /* definitions: 1 */

struct SLAVIO_CPUINTCTLState {
uint32_t                   intreg_pending;       /*     0     4 */

/* XXX 4 bytes hole, try to pack */

struct SLAVIO_INTCTLState * master;              /*     8     8 */
uint32_t                   cpu;                  /*    16     4 */
uint32_t                   irl_out;              /*    20     4 */

/* size: 24, cachelines: 1 */
/* sum members: 20, holes: 1, sum holes: 4 */
/* last cacheline: 24 bytes */
}; /* definitions: 1 */

Fix by rearranging the structures to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

slavio_misc: avoid structure holes spotted by pahole

Report from pahole on amd64 host:
struct MiscState {
SysBusDevice               busdev;               /*     0  5648 */
/* --- cacheline 88 boundary (5632 bytes) was 16 bytes ago --- */
qemu_irq                   irq;                  /*  5648     8 */
uint32_t                   dummy;                /*  5656     4 */
uint8_t                    config;               /*  5660     1 */
uint8_t                    aux1;                 /*  5661     1 */
uint8_t                    aux2;                 /*  5662     1 */
uint8_t                    diag;                 /*  5663     1 */
uint8_t                    mctrl;                /*  5664     1 */
uint8_t                    sysctrl;              /*  5665     1 */
uint16_t                   leds;                 /*  5666     2 */

/* XXX 4 bytes hole, try to pack */

qemu_irq                   fdc_tc;               /*  5672     8 */

/* size: 5680, cachelines: 89 */
/* sum members: 5676, holes: 1, sum holes: 4 */
/* last cacheline: 48 bytes */
}; /* definitions: 1 */

Fix by rearranging the structure to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

slavio_timer: avoid structure holes spotted by pahole

Report from pahole on amd64 host:
struct SLAVIO_TIMERState {
SysBusDevice               busdev;               /*     0  5648 */
/* --- cacheline 88 boundary (5632 bytes) was 16 bytes ago --- */
uint32_t                   num_cpus;             /*  5648     4 */

/* XXX 4 bytes hole, try to pack */

CPUTimerState              cputimer[17];         /*  5656   816 */
/* --- cacheline 101 boundary (6464 bytes) was 8 bytes ago --- */
uint32_t                   cputimer_mode;        /*  6472     4 */

/* size: 6480, cachelines: 102 */
/* sum members: 6472, holes: 1, sum holes: 4 */
/* padding: 4 */
/* last cacheline: 16 bytes */
}; /* definitions: 1 */

struct CPUTimerState {
qemu_irq                   irq;                  /*     0     8 */
ptimer_state *             timer;                /*     8     8 */
uint32_t                   count;                /*    16     4 */
uint32_t                   counthigh;            /*    20     4 */
uint32_t                   reached;              /*    24     4 */

/* XXX 4 bytes hole, try to pack */

uint64_t                   limit;                /*    32     8 */
uint32_t                   running;              /*    40     4 */

/* size: 48, cachelines: 1 */
/* sum members: 40, holes: 1, sum holes: 4 */
/* padding: 4 */
/* last cacheline: 48 bytes */
}; /* definitions: 1 */

Fix by rearranging the structures to avoid padding.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

monitor: Prevent sign-extension of 32-bit addresses printed by info tlb

This is the same fix that was recently applied to info mem.  Before
this change, info tlb output looked like:

ffffffffffffc000: 000000000fffc000 --------W
ffffffffffffd000: 000000000fffd000 --------W
ffffffffffffe000: 000000000fffe000 --------W
fffffffffffff000: 000000000ffff000 --------W

With this change, it looks like

00000000ffffc000: 000000000fffc000 --------W
00000000ffffd000: 000000000fffd000 --------W
00000000ffffe000: 000000000fffe000 --------W
00000000fffff000: 000000000ffff000 --------W

Signed-off-by: Austin Clements <amdragon@mit.edu>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

Merge branch 'queues/slirp' of git://git.kiszka.org/qemu

* 'queues/slirp' of git://git.kiszka.org/qemu:
  slirp: Fix bit field types in IP header structs

tcg-ia64: Fix typos in AREG0 setup in prologue.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

tcg-hppa: Fix CPU_TEMP_BUF_NLONGS oversight.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

tcg: Constant fold neg, andc, orc, eqv, nand, nor.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

tcg: Always define all of the TCGOpcode enum members.

By always defining these symbols, we can eliminate a lot of ifdefs.

To allow this to be checked reliably, the semantics of the
TCG_TARGET_HAS_* macros must be changed from def/undef to true/false.
This allows even more ifdefs to be removed, converting them into
C if statements.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

tcg: Add and use TCG_OPF_64BIT.

This allows the simplification of the op_bits function from
tcg/optimize.c.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

Check for presence of compiler -pthread flag.

OpenBSD / FreeBSD and some other OS's require the use of
cc -pthread to link threaded programs so have QEMU's
configure script check for the presence of the flag
and use it if so.

Signed-off-by: Brad Smith <brad@comstyle.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

Convert last qemu_free and qemu_malloc uses

7267c0947d7e8ae5dff7bafd932c3bc285f43e5c missed
a few cases, fix them.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

monitor: Show combined protection bits in "info mem"

Previously, "info mem" considered and displayed only the last-level
protection bits for a memory range, which doesn't accurrately
represent the protection of that range.  Now it shows the combined
protection.

Signed-off-by: Austin Clements <amdragon@mit.edu>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

monitor: Fix "info mem" to print the last memory range

"info mem" groups its output into contiguous ranges with identical
protection bits, but previously forgot to print the last range.

Signed-off-by: Austin Clements <amdragon@mit.edu>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

monitor: Prevent sign-extension of 32-bit addresses printed by info mem

Previously, on 32-bit i386, info mem used signed 32-bit int's to store
the page table indexes.  As a result, address calculation was done in
32 bits and then incorrectly sign-extended to 64 bits, yielding output
like

ffffffffef000000-ffffffffef031000 0000000000031000 ur-
ffffffffef7bc000-ffffffffef7bd000 0000000000001000 urw
ffffffffef7bd000-ffffffffef7be000 0000000000001000 ur-

This makes these indexes unsigned, which yields correct output

00000000ef000000-00000000ef031000 0000000000031000 ur-
00000000ef7bc000-00000000ef7bd000 0000000000001000 urw
00000000ef7bd000-00000000ef7be000 0000000000001000 ur-

Signed-off-by: Austin Clements <amdragon@mit.edu>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

w32: Fix format string regression

Commit 953ffe0f935f40c0d6061d69e76e0339393b54f8
introduced FMT_pid which is wrong for w32 and w64 getpid():
those getpid() implementations always return an int value.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

w64: Add definition of FMT_pid

For mingw-w64, pid_t is _pid_t which is __int64,
so this platform needs its own definition of FMT_pid.

Reviewed-by: Andreas Färber <andreas.faerber@web.de>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

Fix conversions from pointer to tcg_target_long

tcg_gen_exit_tb takes a parameter of type tcg_target_long,
so the type casts of pointer to long should be replaced by
type casts of pointer to tcg_target_long.

These changes are needed for build environments where
sizeof(long) != sizeof(void *), especially for w64.

See 4b4a72e55660abf7efe85aca78762dcfea5519ad which fixed the
same issue for the other targets.

Cc: Alexander Graf <agraf@suse.de>
Cc: Guan Xuetao <gxt@mprc.pku.edu.cn>
Acked-by: Guan Xuetao<gxt@mprc.pku.edu.cn>
Acked-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

Remove remenants of qemu_malloc

This covers the various check commands

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Update HACKING to refer to g_malloc instead of qemu_malloc

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Add trace points for g_malloc/g_free functions

Derived from a patch submitted by Avi Kivity.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Remove qemu_malloc/qemu_free

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Use glib memory allocation and free functions

qemu_malloc/qemu_free no longer exist after this commit.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Make glib mandatory and fixup utils appropriately

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Reorder default ram_size initialization

code_gen_alloc depends on it, and that is now called earlier via
configure_accelerator.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

scsi-disk: fix DPRINTF

The variable 'status' does not exist anymore, adjust DPRINTF
accordingly.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

target-cris/opcode-cris.h: rename REG_PC/SP to CRIS_REG_PC/SP

The REG_PC constant used in opcode-cris.h can clash with a
similar define in system include files. In particular the
Ubuntu Lucid SPARC signal.h will define REG_PC, and since
qemu-common.h now includes signal.h this was causing compile
failures. Rename the constants to avoid this issue.
(NB that REG_SP is not actually used within QEMU.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

scsi: do not overwrite memory on REQUEST SENSE commands with a large buffer

Other scsi_target_reqops commands were careful about not using r->cmd.xfer
directly, and instead always cap it to a fixed length.  This was not done
for REQUEST SENSE, and this patch fixes it.

Reported-by: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

slirp: Fix bit field types in IP header structs

-mms-bitfields prevents that the bitfields in current IP header structs
are packed into a single byte as it is required. Fix this by using
uint8_t as backing type.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>

hw/scsi-bus.c: Fix use of uninitialised variable

Don't use req before it has been initialised in scsi_req_new().
This fixes a compile failure due to gcc complaining about this.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

qapi: fix build issue due to missing newline in generated header

Fixes a build issue on RHEL5, and potentially other distros, where gcc
will generate an error due to us not writing a trailing "\n" when
generating *qmp-commands.h

Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

memory: correct documentation typos

Noted by Drew Jones.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

memory: add API for creating ROM/device regions

ROM/device regions act as mapped RAM for reads, can I/O memory for
writes.  This allow emulation of flash devices.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

memory: reclaim resources when a memory region is destroyed for good

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

build: Move QEMU_INCLUDES before QEMU_CFLAGS

This patch fixes build when any of the include paths from QEMU_CFLAGS
contains a header file with similar name to a header file in qemu
sources. I hit it with error.h included by qapi/qapi-types-core.h. GCC
decided to use /usr/include/alsa/error.h instead of qemu's error.h.

Tested-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

configure: Disable guest_agent for mingw32

guest_agent is not supported for mingw32, so the default value
should be 'no', not 'yes'.

This removes the dependencies to glib-2.0 and python which
makes native and cross builds for w32 much easier (no need
to get and install these extra packages).

It also avoids the problems caused by different bitfield alignment
which is required by glib-2.0.

It is still possible to set guest_agent=yes via configure option.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: add special traces for common commands

Can be useful when debugging the device scan phase.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: report unit attention on reset

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: add support for unit attention conditions

Unit attention conditions override any sense data the device already
has.  Their signaling and clearing is handled entirely by the SCSIBus
code, and they are completely transparent to the SCSIDevices.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: add a bunch more common sense codes

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: move handling of REQUEST SENSE to common code

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: move handling of REPORT LUNS and invalid LUNs to common code

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: move request parsing to common code

Also introduce the first occurrence of "independent" SCSIReqOps,
to handle invalid commands in common code.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: push lun field to SCSIDevice

This will let SCSIBus detect requests sent to an invalid LUN, and
handle them itself.  However, there will be still support for only one
LUN per target

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: introduce SCSICommand

This struct is currently unnamed.  Give it a name and use it
explicitly to decouple (some parts of) CDB parsing from
SCSIRequest.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: pass cdb already to scsi_req_new

Right now the CDB is not passed to the SCSIBus until scsi_req_enqueue.
Passing it to scsi_req_new will let scsi_req_new dispatch common requests
through different reqops.

Moving the memcpy to scsi_req_new is a hack that will go away as
soon as scsi_req_new will also take care of the parsing.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: move request-related callbacks from SCSIDeviceInfo to SCSIReqOps

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: introduce SCSIReqOps

This will let allow requests to be dispatched through different callbacks,
either common or per-device.

This patch adjusts the API, the next one will move members to SCSIReqOps.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: move sense handling to generic code

With this patch, sense data is stored in the generic data structures
for SCSI devices and requests.  The SCSI layer takes care of storing
sense data in the SCSIDevice for the subsequent REQUEST SENSE command.

At the same time, get_sense is removed and scsi_req_get_sense can use
an entirely generic implementation.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi: pass status when completing

A small improvement in the SCSI request API.  Pass the status
at the time the request is completed, so that we can assert that
no request is completed twice.  This would have detected the
problem fixed in the previous patch.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

vscsi: always use get_sense

vscsi supports autosensing by providing sense data directly in the
response.  When get_sense was added, the older state machine approach
that sent REQUEST SENSE commands separately was left in place.  Remove
it, all existing SCSIDevices do support autosensing and the next patches
will make the support come for free from the SCSIBus.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

scsi-disk: no need to call scsi_req_data on a short read

In fact, if the HBA's transfer_data callback goes on with scsi_req_continue
the request will be completed successfully instead of showing a failure.
It can even cause a segmentation fault.

An easy way to trigger it is "eject -f cd" during installation (during media
test if the installer does something like that).

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

pc: make vgabios exit port more useful

We've always listened on port 501 for vgabios panic messages.  In the entire
time I've worked on QEMU, I've never actually seen a vgabios panic message :-)

If we change the semantics of this port a little bit, it makes it possible to
use it for more interesting use-cases.  I chose this approach instead of adding
a new I/O port because it avoids having a guest visible change.

This change allows single-byte access to port 501 and also uses the value
written to construct an exit code.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Merge remote-tracking branch 'stefanha/trivial-patches' into staging

Merge remote-tracking branch 'aneesh/for-upstream-1' into staging

Merge remote-tracking branch 'spice/spice.v41' into staging

Merge remote-tracking branch 'kraxel/usb.23' into staging

Merge remote-tracking branch 'kraxel/seabios' into staging

fix QLIST usage for RAM list

Spotted while reviewing the migration thread patches.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>

hw/qdev: Don't crash if qdev_create(NULL, ...) fails

If an attempt to create a qdev device on the default sysbus (by passing
NULL as the bus to qdev_create) fails, print a useful error message
rather than crashing trying to dereference a NULL pointer.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>

scsi-bus: use DO_UPCAST

Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>

etrax-ser: printf -> qemu_log.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

etrax: QDevify the Ethernet MAC.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

ui/spice-core: report compiled-version in info spice/query-spice

Signed-off-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

seabios: update to master

commit 8e301472e324b6d6496d8b4ffc66863e99d7a505

user visible changes in seabios:
  * ahci is enabled by default (and thus in this build).
  * bootorder support for ahci.
  * two-pass pci allocator (orders bars by size for better packing).

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

qxl: unbreak after memory API conversion

Break is only noticable with newer spice-server library (0.8.2 release
or 0.9.0 and newer on master branch).

ioport_write's val was changed from uint32_t to uint64_t, this
broke two printfs. Use PRId64 instead of %d.

Signed-off-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

qxl: allowing the command rings to be not empty when spice worker is stopped RHBZ #728984

same as 8927cfbba232e28304734f7afd463c1b84134031, but for qxl_check_state, that was
triggered by qxl_pre_load (which calls qxl_hard_reset, which calls qxl_soft_reset),
and caused the migration target to crash.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

usb-hid: remove usb_hid_datain_cb

No users left, all migrated over to hw/hid.[ch].
Yea!  Zap it!

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

milkymist-softusb: use hid code directly

Remove the dummy USB device and use the HID code directly. Use the HID code
for the mouse support, too.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

usb-hid: use hid vmstate macro

Use new hid vmstate macro. Version stays the same, because there is no
reordering of the fields.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

hid: introduce hid vmstate macros

Add VMSTATE macros to describe a HIDState. Based on usb-hid.c descriptions.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

hid: register kbd hander in init()

Register the keyboard event handler in hid's init() instead of its reset()
function.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

usb/hid: add hid_pointer_activate, use it

HID reorganziation broke the usb tablet in windows xp.  The reason is
that xp activates idle before it starts polling, which creates a
chicken-and-egg issue:  We don't call hid_pointer_poll because there are
no pending events.  We don't get any events because the activation code
in hid_pointer_poll is never executed and thus all pointer events are
routed to the PS/2 mouse by qemu.

Fix this by creating a hid_pointer_activate function and call it from
usb-hid when the guest sets the idle state.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

Merge remote-tracking branch 'pm-arm/for-upstream' into pm

etrax: Allocate DMA connections at board level.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

etrax: Remove hw/etraxfs.c.

The Bare ETRAX FS board was a fictive machine that I used when
developing the CRIS system emulation. Since we support the
real AXIS-dev88 developer boards, there is no reason to
keep the fictive one around.

This commit also removes the double registration of the axis-dev88
board.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

lance: unbreak after memory API conversion

The conversion passed the wrong opaque pointer, causing a crash on first use.
Pass the correct opaque.

Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

Merge remote-tracking branch 'qemu-kvm/uq/master' into staging

hw/9pfs: Update vfs_rename to use coroutines

I guess TRENAME 9p operation needs an update. The 9p op should
more similar renameat. Otherwise anything other than path cannot track
the fid.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

hw/9pfs: Add yeild support to rename coroutine

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

hw/9pfs: Update v9fs_remove to use coroutines

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

hw/9pfs: Add yield support for remove

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

hw/9pfs: Update mkdir to use coroutines

Signed-off-by: Venkateswararao Jujjuri "<jvrao@linux.vnet.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

hw/9pfs: Add yield support for mkdir coroutine

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

hw/9pfs: Update v9fs_mknod to use coroutines

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

hw/9pfs: Add yield support to mknod coroutine

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

hw/9pfs: Update v9fs_xattrcreate to use coroutines

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>