June 2018 - Wine-Devel - list.winehq.org

[PATCH 3/3] winedbg: Always start the interactive debugger in 64-bit mode if in a 64-bit prefix.

by Zebediah Figura

Fixes https://bugs.winehq.org/show_bug.cgi?id=45109 Signed-off-by: Zebediah Figura <zfigura(a)codeweavers.com> --- programs/winedbg/winedbg.c | 32 ++++++++++++++++++++++++++++++++ 1 file changed, 32 insertions(+) diff --git a/programs/winedbg/winedbg.c b/programs/winedbg/winedbg.c index 1ac4187..0e45186 100644 --- a/programs/winedbg/winedbg.c +++ b/programs/winedbg/winedbg.c @@ -632,6 +632,35 @@ static LONG CALLBACK top_filter( EXCEPTION_POINTERS *ptr ) return EXCEPTION_EXECUTE_HANDLER; } +static void restart_if_wow64(void) +{ + BOOL is_wow64; + + if (IsWow64Process( GetCurrentProcess(), &is_wow64 ) && is_wow64) + { + STARTUPINFOW si; + PROCESS_INFORMATION pi; + WCHAR filename[MAX_PATH]; + void *redir; + DWORD exit_code; + + memset( &si, 0, sizeof(si) ); + si.cb = sizeof(si); + GetModuleFileNameW( 0, filename, MAX_PATH ); + + Wow64DisableWow64FsRedirection( &redir ); + if (CreateProcessW( filename, GetCommandLineW(), NULL, NULL, FALSE, 0, NULL, NULL, &si, &pi )) + { + WINE_TRACE( "restarting %s\n", wine_dbgstr_w(filename) ); + WaitForSingleObject( pi.hProcess, INFINITE ); + GetExitCodeProcess( pi.hProcess, &exit_code ); + ExitProcess( exit_code ); + } + else WINE_ERR( "failed to restart 64-bit %s, err %d\n", wine_dbgstr_w(filename), GetLastError() ); + Wow64RevertWow64FsRedirection( redir ); + } +} + int main(int argc, char** argv) { int retv = 0; @@ -654,6 +683,7 @@ int main(int argc, char** argv) if (argc && !strcmp(argv[0], "--gdb")) { + restart_if_wow64(); retv = gdb_main(argc, argv); if (retv == -1) dbg_winedbg_usage(FALSE); return retv; @@ -727,6 +757,8 @@ int main(int argc, char** argv) case start_error_init: return -1; } + restart_if_wow64(); + dbg_start_interactive(hFile); return 0; -- 2.7.4

June 13, 2018

[PATCH 2/3] winedbg/gdbproxy: Store the register map inside the be_cpu struct.

by Zebediah Figura

Signed-off-by: Zebediah Figura <zfigura(a)codeweavers.com> --- programs/winedbg/be_arm.c | 24 +++ programs/winedbg/be_arm64.c | 41 ++++++ programs/winedbg/be_cpu.h | 10 ++ programs/winedbg/be_i386.c | 48 ++++++ programs/winedbg/be_ppc.c | 80 ++++++++++ programs/winedbg/be_x86_64.c | 64 ++++++++ programs/winedbg/debugger.h | 2 + programs/winedbg/gdbproxy.c | 343 +++++-------------------------------------- 8 files changed, 306 insertions(+), 306 deletions(-) diff --git a/programs/winedbg/be_arm.c b/programs/winedbg/be_arm.c index 9883476..0c65757 100644 --- a/programs/winedbg/be_arm.c +++ b/programs/winedbg/be_arm.c @@ -1904,6 +1904,28 @@ static BOOL be_arm_set_context(HANDLE thread, const dbg_ctx_t *ctx) return SetThreadContext(thread, &ctx->ctx); } +#define REG(r,gs) {FIELD_OFFSET(CONTEXT, r), sizeof(((CONTEXT*)NULL)->r), gs} + +static struct gdb_register be_arm_gdb_register_map[] = { + REG(R0, 4), + REG(R1, 4), + REG(R2, 4), + REG(R3, 4), + REG(R4, 4), + REG(R5, 4), + REG(R6, 4), + REG(R7, 4), + REG(R8, 4), + REG(R9, 4), + REG(R10, 4), + REG(R11, 4), + REG(R12, 4), + REG(Sp, 4), + REG(Lr, 4), + REG(Pc, 4), + REG(Cpsr, 4), +}; + struct backend_cpu be_arm = { IMAGE_FILE_MACHINE_ARMNT, @@ -1931,5 +1953,7 @@ struct backend_cpu be_arm = be_arm_fetch_float, be_arm_store_integer, be_arm_get_context, + be_arm_gdb_register_map, + ARRAY_SIZE(be_arm_gdb_register_map), }; #endif diff --git a/programs/winedbg/be_arm64.c b/programs/winedbg/be_arm64.c index 418d6e0..40ef96b 100644 --- a/programs/winedbg/be_arm64.c +++ b/programs/winedbg/be_arm64.c @@ -293,6 +293,45 @@ static BOOL be_arm64_set_context(HANDLE thread, const dbg_ctx_t *ctx) return SetThreadContext(thread, &ctx->ctx); } +#define REG(r,gs) {FIELD_OFFSET(CONTEXT, r), sizeof(((CONTEXT*)NULL)->r), gs} + +static struct gdb_register be_arm64_gdb_register_map[] = { + REG(Cpsr, 4), + REG(u.s.X0, 8), + REG(u.s.X1, 8), + REG(u.s.X2, 8), + REG(u.s.X3, 8), + REG(u.s.X4, 8), + REG(u.s.X5, 8), + REG(u.s.X6, 8), + REG(u.s.X7, 8), + REG(u.s.X8, 8), + REG(u.s.X9, 8), + REG(u.s.X10, 8), + REG(u.s.X11, 8), + REG(u.s.X12, 8), + REG(u.s.X13, 8), + REG(u.s.X14, 8), + REG(u.s.X15, 8), + REG(u.s.X16, 8), + REG(u.s.X17, 8), + REG(u.s.X18, 8), + REG(u.s.X19, 8), + REG(u.s.X20, 8), + REG(u.s.X21, 8), + REG(u.s.X22, 8), + REG(u.s.X23, 8), + REG(u.s.X24, 8), + REG(u.s.X25, 8), + REG(u.s.X26, 8), + REG(u.s.X27, 8), + REG(u.s.X28, 8), + REG(u.s.Fp, 8), + REG(u.s.Lr, 8), + REG(Sp, 8), + REG(Pc, 8), +}; + struct backend_cpu be_arm64 = { IMAGE_FILE_MACHINE_ARM64, @@ -320,5 +359,7 @@ struct backend_cpu be_arm64 = be_arm64_fetch_float, be_arm64_store_integer, be_arm64_get_context, + be_arm64_gdb_register_map, + ARRAY_SIZE(be_arm64_gdb_register_map), }; #endif diff --git a/programs/winedbg/be_cpu.h b/programs/winedbg/be_cpu.h index 1a67c06..ea559b4 100644 --- a/programs/winedbg/be_cpu.h +++ b/programs/winedbg/be_cpu.h @@ -22,6 +22,13 @@ enum be_cpu_addr {be_cpu_addr_pc, be_cpu_addr_stack, be_cpu_addr_frame}; enum be_xpoint_type {be_xpoint_break, be_xpoint_watch_exec, be_xpoint_watch_read, be_xpoint_watch_write, be_xpoint_free=-1}; +struct gdb_register +{ + size_t ctx_offset; + size_t ctx_length; + size_t gdb_length; +}; + struct backend_cpu { const DWORD machine; @@ -118,6 +125,9 @@ struct backend_cpu BOOL (*get_context)(HANDLE thread, dbg_ctx_t *ctx); BOOL (*set_context)(HANDLE thread, const dbg_ctx_t *ctx); + + const struct gdb_register *gdb_register_map; + const size_t gdb_num_regs; }; /* some handy functions for non segmented CPUs */ diff --git a/programs/winedbg/be_i386.c b/programs/winedbg/be_i386.c index 4e279ec..3a2b91a 100644 --- a/programs/winedbg/be_i386.c +++ b/programs/winedbg/be_i386.c @@ -866,6 +866,52 @@ static BOOL be_i386_set_context(HANDLE thread, const dbg_ctx_t *ctx) return Wow64SetThreadContext(thread, &ctx->x86); } +#define REG(r,gs) {FIELD_OFFSET(WOW64_CONTEXT, r), sizeof(((WOW64_CONTEXT*)NULL)->r), gs} + +static struct gdb_register be_i386_gdb_register_map[] = { + REG(Eax, 4), + REG(Ecx, 4), + REG(Edx, 4), + REG(Ebx, 4), + REG(Esp, 4), + REG(Ebp, 4), + REG(Esi, 4), + REG(Edi, 4), + REG(Eip, 4), + REG(EFlags, 4), + REG(SegCs, 4), + REG(SegSs, 4), + REG(SegDs, 4), + REG(SegEs, 4), + REG(SegFs, 4), + REG(SegGs, 4), + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.RegisterArea[ 0]), 10, 10 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.RegisterArea[10]), 10, 10 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.RegisterArea[20]), 10, 10 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.RegisterArea[30]), 10, 10 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.RegisterArea[40]), 10, 10 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.RegisterArea[50]), 10, 10 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.RegisterArea[60]), 10, 10 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.RegisterArea[70]), 10, 10 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.ControlWord), 2, 4 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.StatusWord), 2, 4 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.TagWord), 2, 4 }, + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.ErrorSelector), 2, 4 }, + REG(FloatSave.ErrorOffset, 4 ), + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.DataSelector), 2, 4 }, + REG(FloatSave.DataOffset, 4 ), + { FIELD_OFFSET(WOW64_CONTEXT, FloatSave.ErrorSelector)+2, 2, 4 }, + { FIELD_OFFSET(WOW64_CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[0]), 16, 16 }, + { FIELD_OFFSET(WOW64_CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[1]), 16, 16 }, + { FIELD_OFFSET(WOW64_CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[2]), 16, 16 }, + { FIELD_OFFSET(WOW64_CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[3]), 16, 16 }, + { FIELD_OFFSET(WOW64_CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[4]), 16, 16 }, + { FIELD_OFFSET(WOW64_CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[5]), 16, 16 }, + { FIELD_OFFSET(WOW64_CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[6]), 16, 16 }, + { FIELD_OFFSET(WOW64_CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[7]), 16, 16 }, + { FIELD_OFFSET(WOW64_CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, MxCsr), 4, 4 }, +}; + struct backend_cpu be_i386 = { IMAGE_FILE_MACHINE_I386, @@ -894,5 +940,7 @@ struct backend_cpu be_i386 = be_i386_store_integer, be_i386_get_context, be_i386_set_context, + be_i386_gdb_register_map, + ARRAY_SIZE(be_i386_gdb_register_map), }; #endif diff --git a/programs/winedbg/be_ppc.c b/programs/winedbg/be_ppc.c index 40aee55..b0e6f92 100644 --- a/programs/winedbg/be_ppc.c +++ b/programs/winedbg/be_ppc.c @@ -195,6 +195,84 @@ static BOOL be_ppc_set_context(HANDLE thread, const dbg_ctx_t *ctx) return SetThreadContext(thread, &ctx->ctx); } +#define REG(r,gs,m) {FIELD_OFFSET(CONTEXT, r), sizeof(((CONTEXT*)NULL)->r), gs, m} + +static struct gdb_register be_ppc_gdb_register_map[] = { + REG(Gpr0, 4), + REG(Gpr1, 4), + REG(Gpr2, 4), + REG(Gpr3, 4), + REG(Gpr4, 4), + REG(Gpr5, 4), + REG(Gpr6, 4), + REG(Gpr7, 4), + REG(Gpr8, 4), + REG(Gpr9, 4), + REG(Gpr10, 4), + REG(Gpr11, 4), + REG(Gpr12, 4), + REG(Gpr13, 4), + REG(Gpr14, 4), + REG(Gpr15, 4), + REG(Gpr16, 4), + REG(Gpr17, 4), + REG(Gpr18, 4), + REG(Gpr19, 4), + REG(Gpr20, 4), + REG(Gpr21, 4), + REG(Gpr22, 4), + REG(Gpr23, 4), + REG(Gpr24, 4), + REG(Gpr25, 4), + REG(Gpr26, 4), + REG(Gpr27, 4), + REG(Gpr28, 4), + REG(Gpr29, 4), + REG(Gpr30, 4), + REG(Gpr31, 4), + REG(Fpr0, 4), + REG(Fpr1, 4), + REG(Fpr2, 4), + REG(Fpr3, 4), + REG(Fpr4, 4), + REG(Fpr5, 4), + REG(Fpr6, 4), + REG(Fpr7, 4), + REG(Fpr8, 4), + REG(Fpr9, 4), + REG(Fpr10, 4), + REG(Fpr11, 4), + REG(Fpr12, 4), + REG(Fpr13, 4), + REG(Fpr14, 4), + REG(Fpr15, 4), + REG(Fpr16, 4), + REG(Fpr17, 4), + REG(Fpr18, 4), + REG(Fpr19, 4), + REG(Fpr20, 4), + REG(Fpr21, 4), + REG(Fpr22, 4), + REG(Fpr23, 4), + REG(Fpr24, 4), + REG(Fpr25, 4), + REG(Fpr26, 4), + REG(Fpr27, 4), + REG(Fpr28, 4), + REG(Fpr29, 4), + REG(Fpr30, 4), + REG(Fpr31, 4), + + REG(Iar, 4), + REG(Msr, 4), + REG(Cr, 4), + REG(Lr, 4), + REG(Ctr, 4), + REG(Xer, 4), + /* FIXME: MQ is missing? FIELD_OFFSET(CONTEXT, Mq), */ + /* see gdb/nlm/ppc.c */ +}; + struct backend_cpu be_ppc = { IMAGE_FILE_MACHINE_POWERPC, @@ -223,5 +301,7 @@ struct backend_cpu be_ppc = be_ppc_store_integer, be_ppc_get_context, be_ppc_set_context, + be_ppc_gdb_register_map, + ARRAY_SIZE(be_ppc_gdb_register_map), }; #endif diff --git a/programs/winedbg/be_x86_64.c b/programs/winedbg/be_x86_64.c index 15f84a6..b7f75c6 100644 --- a/programs/winedbg/be_x86_64.c +++ b/programs/winedbg/be_x86_64.c @@ -689,6 +689,68 @@ static BOOL be_x86_64_set_context(HANDLE thread, const dbg_ctx_t *ctx) return SetThreadContext(thread, &ctx->ctx); } +#define REG(r,gs) {FIELD_OFFSET(CONTEXT, r), sizeof(((CONTEXT*)NULL)->r), gs} + +static struct gdb_register be_x86_64_gdb_register_map[] = { + REG(Rax, 8), + REG(Rbx, 8), + REG(Rcx, 8), + REG(Rdx, 8), + REG(Rsi, 8), + REG(Rdi, 8), + REG(Rbp, 8), + REG(Rsp, 8), + REG(R8, 8), + REG(R9, 8), + REG(R10, 8), + REG(R11, 8), + REG(R12, 8), + REG(R13, 8), + REG(R14, 8), + REG(R15, 8), + REG(Rip, 8), + REG(EFlags, 4), + REG(SegCs, 4), + REG(SegSs, 4), + REG(SegDs, 4), + REG(SegEs, 4), + REG(SegFs, 4), + REG(SegGs, 4), + { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 0]), 10, 10 }, + { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 1]), 10, 10 }, + { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 2]), 10, 10 }, + { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 3]), 10, 10 }, + { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 4]), 10, 10 }, + { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 5]), 10, 10 }, + { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 6]), 10, 10 }, + { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 7]), 10, 10 }, + REG(u.FltSave.ControlWord, 4), + REG(u.FltSave.StatusWord, 4), + REG(u.FltSave.TagWord, 4), + REG(u.FltSave.ErrorSelector, 4), + REG(u.FltSave.ErrorOffset, 4), + REG(u.FltSave.DataSelector, 4), + REG(u.FltSave.DataOffset, 4), + REG(u.FltSave.ErrorOpcode, 4), + REG(u.s.Xmm0, 16), + REG(u.s.Xmm1, 16), + REG(u.s.Xmm2, 16), + REG(u.s.Xmm3, 16), + REG(u.s.Xmm4, 16), + REG(u.s.Xmm5, 16), + REG(u.s.Xmm6, 16), + REG(u.s.Xmm7, 16), + REG(u.s.Xmm8, 16), + REG(u.s.Xmm9, 16), + REG(u.s.Xmm10, 16), + REG(u.s.Xmm11, 16), + REG(u.s.Xmm12, 16), + REG(u.s.Xmm13, 16), + REG(u.s.Xmm14, 16), + REG(u.s.Xmm15, 16), + REG(u.FltSave.MxCsr, 4), +}; + struct backend_cpu be_x86_64 = { IMAGE_FILE_MACHINE_AMD64, @@ -717,5 +779,7 @@ struct backend_cpu be_x86_64 = be_x86_64_store_integer, be_x86_64_get_context, be_x86_64_set_context, + be_x86_64_gdb_register_map, + ARRAY_SIZE(be_x86_64_gdb_register_map), }; #endif diff --git a/programs/winedbg/debugger.h b/programs/winedbg/debugger.h index 490bd06..3fc8daa 100644 --- a/programs/winedbg/debugger.h +++ b/programs/winedbg/debugger.h @@ -39,6 +39,8 @@ #include "oaidl.h" #include <wine/list.h> +#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) + #define ADDRSIZE (dbg_curr_process->be_cpu->pointer_size) #define ADDRWIDTH (ADDRSIZE * 2) diff --git a/programs/winedbg/gdbproxy.c b/programs/winedbg/gdbproxy.c index 81cd7de..869d56b 100644 --- a/programs/winedbg/gdbproxy.c +++ b/programs/winedbg/gdbproxy.c @@ -183,233 +183,12 @@ static unsigned char checksum(const char* ptr, int len) return cksum; } -/* =============================================== * - * C P U H A N D L E R S * - * =============================================== * - */ - -/* This struct helps us to manage the different representations of a register: - * ctx_offset and ctx_length are the location and size in Win32 CONTEXT - * gdb_length is the length gdb expects on the wire - * As the two sizes could be different, we have to convert between the two - * (for example, on x86_64, Seg?s are 4 bytes on the wire and 2 in CONTEXT) - */ -struct cpu_register -{ - size_t ctx_offset; - size_t ctx_length; - size_t gdb_length; -}; - -#define REG(r,gs) {FIELD_OFFSET(CONTEXT, r), sizeof(((CONTEXT*)NULL)->r), gs} - #ifdef __i386__ -typedef struct DECLSPEC_ALIGN(16) _M128A { - ULONGLONG Low; - LONGLONG High; -} M128A, *PM128A; - -typedef struct _XMM_SAVE_AREA32 { - WORD ControlWord; /* 000 */ - WORD StatusWord; /* 002 */ - BYTE TagWord; /* 004 */ - BYTE Reserved1; /* 005 */ - WORD ErrorOpcode; /* 006 */ - DWORD ErrorOffset; /* 008 */ - WORD ErrorSelector; /* 00c */ - WORD Reserved2; /* 00e */ - DWORD DataOffset; /* 010 */ - WORD DataSelector; /* 014 */ - WORD Reserved3; /* 016 */ - DWORD MxCsr; /* 018 */ - DWORD MxCsr_Mask; /* 01c */ - M128A FloatRegisters[8]; /* 020 */ - M128A XmmRegisters[16]; /* 0a0 */ - BYTE Reserved4[96]; /* 1a0 */ -} XMM_SAVE_AREA32, *PXMM_SAVE_AREA32; - static const char target_xml[] = ""; -static struct cpu_register cpu_register_map[] = { - REG(Eax, 4), - REG(Ecx, 4), - REG(Edx, 4), - REG(Ebx, 4), - REG(Esp, 4), - REG(Ebp, 4), - REG(Esi, 4), - REG(Edi, 4), - REG(Eip, 4), - REG(EFlags, 4), - REG(SegCs, 4), - REG(SegSs, 4), - REG(SegDs, 4), - REG(SegEs, 4), - REG(SegFs, 4), - REG(SegGs, 4), - { FIELD_OFFSET(CONTEXT, FloatSave.RegisterArea[ 0]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, FloatSave.RegisterArea[10]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, FloatSave.RegisterArea[20]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, FloatSave.RegisterArea[30]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, FloatSave.RegisterArea[40]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, FloatSave.RegisterArea[50]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, FloatSave.RegisterArea[60]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, FloatSave.RegisterArea[70]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, FloatSave.ControlWord), 2, 4 }, - { FIELD_OFFSET(CONTEXT, FloatSave.StatusWord), 2, 4 }, - { FIELD_OFFSET(CONTEXT, FloatSave.TagWord), 2, 4 }, - { FIELD_OFFSET(CONTEXT, FloatSave.ErrorSelector), 2, 4 }, - REG(FloatSave.ErrorOffset, 4 ), - { FIELD_OFFSET(CONTEXT, FloatSave.DataSelector), 2, 4 }, - REG(FloatSave.DataOffset, 4 ), - { FIELD_OFFSET(CONTEXT, FloatSave.ErrorSelector)+2, 2, 4 }, - { FIELD_OFFSET(CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[0]), 16, 16 }, - { FIELD_OFFSET(CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[1]), 16, 16 }, - { FIELD_OFFSET(CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[2]), 16, 16 }, - { FIELD_OFFSET(CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[3]), 16, 16 }, - { FIELD_OFFSET(CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[4]), 16, 16 }, - { FIELD_OFFSET(CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[5]), 16, 16 }, - { FIELD_OFFSET(CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[6]), 16, 16 }, - { FIELD_OFFSET(CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, XmmRegisters[7]), 16, 16 }, - { FIELD_OFFSET(CONTEXT, ExtendedRegisters) + FIELD_OFFSET(XMM_SAVE_AREA32, MxCsr), 4, 4 }, -}; #elif defined(__powerpc__) static const char target_xml[] = ""; -static struct cpu_register cpu_register_map[] = { - REG(Gpr0, 4), - REG(Gpr1, 4), - REG(Gpr2, 4), - REG(Gpr3, 4), - REG(Gpr4, 4), - REG(Gpr5, 4), - REG(Gpr6, 4), - REG(Gpr7, 4), - REG(Gpr8, 4), - REG(Gpr9, 4), - REG(Gpr10, 4), - REG(Gpr11, 4), - REG(Gpr12, 4), - REG(Gpr13, 4), - REG(Gpr14, 4), - REG(Gpr15, 4), - REG(Gpr16, 4), - REG(Gpr17, 4), - REG(Gpr18, 4), - REG(Gpr19, 4), - REG(Gpr20, 4), - REG(Gpr21, 4), - REG(Gpr22, 4), - REG(Gpr23, 4), - REG(Gpr24, 4), - REG(Gpr25, 4), - REG(Gpr26, 4), - REG(Gpr27, 4), - REG(Gpr28, 4), - REG(Gpr29, 4), - REG(Gpr30, 4), - REG(Gpr31, 4), - REG(Fpr0, 4), - REG(Fpr1, 4), - REG(Fpr2, 4), - REG(Fpr3, 4), - REG(Fpr4, 4), - REG(Fpr5, 4), - REG(Fpr6, 4), - REG(Fpr7, 4), - REG(Fpr8, 4), - REG(Fpr9, 4), - REG(Fpr10, 4), - REG(Fpr11, 4), - REG(Fpr12, 4), - REG(Fpr13, 4), - REG(Fpr14, 4), - REG(Fpr15, 4), - REG(Fpr16, 4), - REG(Fpr17, 4), - REG(Fpr18, 4), - REG(Fpr19, 4), - REG(Fpr20, 4), - REG(Fpr21, 4), - REG(Fpr22, 4), - REG(Fpr23, 4), - REG(Fpr24, 4), - REG(Fpr25, 4), - REG(Fpr26, 4), - REG(Fpr27, 4), - REG(Fpr28, 4), - REG(Fpr29, 4), - REG(Fpr30, 4), - REG(Fpr31, 4), - - REG(Iar, 4), - REG(Msr, 4), - REG(Cr, 4), - REG(Lr, 4), - REG(Ctr, 4), - REG(Xer, 4), - /* FIXME: MQ is missing? FIELD_OFFSET(CONTEXT, Mq), */ - /* see gdb/nlm/ppc.c */ -}; #elif defined(__x86_64__) static const char target_xml[] = ""; -static struct cpu_register cpu_register_map[] = { - REG(Rax, 8), - REG(Rbx, 8), - REG(Rcx, 8), - REG(Rdx, 8), - REG(Rsi, 8), - REG(Rdi, 8), - REG(Rbp, 8), - REG(Rsp, 8), - REG(R8, 8), - REG(R9, 8), - REG(R10, 8), - REG(R11, 8), - REG(R12, 8), - REG(R13, 8), - REG(R14, 8), - REG(R15, 8), - REG(Rip, 8), - REG(EFlags, 4), - REG(SegCs, 4), - REG(SegSs, 4), - REG(SegDs, 4), - REG(SegEs, 4), - REG(SegFs, 4), - REG(SegGs, 4), - { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 0]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 1]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 2]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 3]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 4]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 5]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 6]), 10, 10 }, - { FIELD_OFFSET(CONTEXT, u.FltSave.FloatRegisters[ 7]), 10, 10 }, - REG(u.FltSave.ControlWord, 4), - REG(u.FltSave.StatusWord, 4), - REG(u.FltSave.TagWord, 4), - REG(u.FltSave.ErrorSelector, 4), - REG(u.FltSave.ErrorOffset, 4), - REG(u.FltSave.DataSelector, 4), - REG(u.FltSave.DataOffset, 4), - REG(u.FltSave.ErrorOpcode, 4), - REG(u.s.Xmm0, 16), - REG(u.s.Xmm1, 16), - REG(u.s.Xmm2, 16), - REG(u.s.Xmm3, 16), - REG(u.s.Xmm4, 16), - REG(u.s.Xmm5, 16), - REG(u.s.Xmm6, 16), - REG(u.s.Xmm7, 16), - REG(u.s.Xmm8, 16), - REG(u.s.Xmm9, 16), - REG(u.s.Xmm10, 16), - REG(u.s.Xmm11, 16), - REG(u.s.Xmm12, 16), - REG(u.s.Xmm13, 16), - REG(u.s.Xmm14, 16), - REG(u.s.Xmm15, 16), - REG(u.FltSave.MxCsr, 4), -}; #elif defined(__arm__) static const char target_xml[] = "l <target><architecture>arm</architecture>\n" @@ -432,96 +211,43 @@ static const char target_xml[] = " <reg name=\"pc\" bitsize=\"32\" type=\"code_ptr\"/>\n" " <reg name=\"cpsr\" bitsize=\"32\"/>\n" "</feature></target>\n"; - -static struct cpu_register cpu_register_map[] = { - REG(R0, 4), - REG(R1, 4), - REG(R2, 4), - REG(R3, 4), - REG(R4, 4), - REG(R5, 4), - REG(R6, 4), - REG(R7, 4), - REG(R8, 4), - REG(R9, 4), - REG(R10, 4), - REG(R11, 4), - REG(R12, 4), - REG(Sp, 4), - REG(Lr, 4), - REG(Pc, 4), - REG(Cpsr, 4), -}; #elif defined(__aarch64__) static const char target_xml[] = ""; -static struct cpu_register cpu_register_map[] = { - REG(Cpsr, 4), - REG(u.s.X0, 8), - REG(u.s.X1, 8), - REG(u.s.X2, 8), - REG(u.s.X3, 8), - REG(u.s.X4, 8), - REG(u.s.X5, 8), - REG(u.s.X6, 8), - REG(u.s.X7, 8), - REG(u.s.X8, 8), - REG(u.s.X9, 8), - REG(u.s.X10, 8), - REG(u.s.X11, 8), - REG(u.s.X12, 8), - REG(u.s.X13, 8), - REG(u.s.X14, 8), - REG(u.s.X15, 8), - REG(u.s.X16, 8), - REG(u.s.X17, 8), - REG(u.s.X18, 8), - REG(u.s.X19, 8), - REG(u.s.X20, 8), - REG(u.s.X21, 8), - REG(u.s.X22, 8), - REG(u.s.X23, 8), - REG(u.s.X24, 8), - REG(u.s.X25, 8), - REG(u.s.X26, 8), - REG(u.s.X27, 8), - REG(u.s.X28, 8), - REG(u.s.Fp, 8), - REG(u.s.Lr, 8), - REG(Sp, 8), - REG(Pc, 8), -}; #else # error Define the registers map for your CPU #endif -#undef REG -static const size_t cpu_num_regs = (sizeof(cpu_register_map) / sizeof(cpu_register_map[0])); - -static inline void* cpu_register_ptr(dbg_ctx_t *ctx, unsigned idx) +static inline void* cpu_register_ptr(struct gdb_context *gdbctx, + dbg_ctx_t *ctx, unsigned idx) { - assert(idx < cpu_num_regs); - return (char*)ctx + cpu_register_map[idx].ctx_offset; + assert(idx < gdbctx->process->be_cpu->gdb_num_regs); + return (char*)ctx + gdbctx->process->be_cpu->gdb_register_map[idx].ctx_offset; } -static inline DWORD64 cpu_register(dbg_ctx_t *ctx, unsigned idx) +static inline DWORD64 cpu_register(struct gdb_context *gdbctx, + dbg_ctx_t *ctx, unsigned idx) { - switch (cpu_register_map[idx].ctx_length) + switch (gdbctx->process->be_cpu->gdb_register_map[idx].ctx_length) { - case 1: return *(BYTE*)cpu_register_ptr(ctx, idx); - case 2: return *(WORD*)cpu_register_ptr(ctx, idx); - case 4: return *(DWORD*)cpu_register_ptr(ctx, idx); - case 8: return *(DWORD64*)cpu_register_ptr(ctx, idx); + case 1: return *(BYTE*)cpu_register_ptr(gdbctx, ctx, idx); + case 2: return *(WORD*)cpu_register_ptr(gdbctx, ctx, idx); + case 4: return *(DWORD*)cpu_register_ptr(gdbctx, ctx, idx); + case 8: return *(DWORD64*)cpu_register_ptr(gdbctx, ctx, idx); default: - fprintf(stderr, "got unexpected size: %u\n", (unsigned)cpu_register_map[idx].ctx_length); + fprintf(stderr, "got unexpected size: %u\n", + (unsigned)gdbctx->process->be_cpu->gdb_register_map[idx].ctx_length); assert(0); return 0; } } -static inline void cpu_register_hex_from(dbg_ctx_t* ctx, unsigned idx, const char **phex) +static inline void cpu_register_hex_from(struct gdb_context *gdbctx, + dbg_ctx_t* ctx, unsigned idx, const char **phex) { + const struct gdb_register *cpu_register_map = gdbctx->process->be_cpu->gdb_register_map; + if (cpu_register_map[idx].gdb_length == cpu_register_map[idx].ctx_length) - hex_from(cpu_register_ptr(ctx, idx), *phex, cpu_register_map[idx].gdb_length); + hex_from(cpu_register_ptr(gdbctx, ctx, idx), *phex, cpu_register_map[idx].gdb_length); else { DWORD64 val = 0; @@ -536,10 +262,10 @@ static inline void cpu_register_hex_from(dbg_ctx_t* ctx, unsigned idx, const cha } switch (cpu_register_map[idx].ctx_length) { - case 1: *(BYTE*)cpu_register_ptr(ctx, idx) = (BYTE)val; break; - case 2: *(WORD*)cpu_register_ptr(ctx, idx) = (WORD)val; break; - case 4: *(DWORD*)cpu_register_ptr(ctx, idx) = (DWORD)val; break; - case 8: *(DWORD64*)cpu_register_ptr(ctx, idx) = val; break; + case 1: *(BYTE*)cpu_register_ptr(gdbctx, ctx, idx) = (BYTE)val; break; + case 2: *(WORD*)cpu_register_ptr(gdbctx, ctx, idx) = (WORD)val; break; + case 4: *(DWORD*)cpu_register_ptr(gdbctx, ctx, idx) = (DWORD)val; break; + case 8: *(DWORD64*)cpu_register_ptr(gdbctx, ctx, idx) = val; break; default: assert(0); } } @@ -1078,11 +804,14 @@ static enum packet_return packet_reply_error(struct gdb_context* gdbctx, int err static inline void packet_reply_register_hex_to(struct gdb_context* gdbctx, unsigned idx) { + const struct gdb_register *cpu_register_map = gdbctx->process->be_cpu->gdb_register_map; + if (cpu_register_map[idx].gdb_length == cpu_register_map[idx].ctx_length) - packet_reply_hex_to(gdbctx, cpu_register_ptr(&gdbctx->context, idx), cpu_register_map[idx].gdb_length); + packet_reply_hex_to(gdbctx, cpu_register_ptr(gdbctx, &gdbctx->context, idx), + cpu_register_map[idx].gdb_length); else { - DWORD64 val = cpu_register(&gdbctx->context, idx); + DWORD64 val = cpu_register(gdbctx, &gdbctx->context, idx); unsigned i; for (i = 0; i < cpu_register_map[idx].gdb_length; i++) @@ -1117,7 +846,7 @@ static enum packet_return packet_reply_status(struct gdb_context* gdbctx) packet_reply_val(gdbctx, dbg_curr_thread->tid, 4); packet_reply_catc(gdbctx, ';'); - for (i = 0; i < cpu_num_regs; i++) + for (i = 0; i < gdbctx->process->be_cpu->gdb_num_regs; i++) { /* FIXME: this call will also grow the buffer... * unneeded, but not harmful @@ -1453,14 +1182,16 @@ static enum packet_return packet_read_registers(struct gdb_context* gdbctx) } packet_reply_open(gdbctx); - for (i = 0; i < cpu_num_regs; i++) + for (i = 0; i < gdbctx->process->be_cpu->gdb_num_regs; i++) packet_reply_register_hex_to(gdbctx, i); + packet_reply_close(gdbctx); return packet_done; } static enum packet_return packet_write_registers(struct gdb_context* gdbctx) { + const size_t cpu_num_regs = gdbctx->process->be_cpu->gdb_num_regs; unsigned i; dbg_ctx_t ctx; dbg_ctx_t *pctx = &gdbctx->context; @@ -1476,7 +1207,7 @@ static enum packet_return packet_write_registers(struct gdb_context* gdbctx) ptr = gdbctx->in_packet; for (i = 0; i < cpu_num_regs; i++) - cpu_register_hex_from(pctx, i, &ptr); + cpu_register_hex_from(gdbctx, pctx, i, &ptr); if (pctx != &gdbctx->context && !gdbctx->process->be_cpu->set_context(gdbctx->other_thread->handle, pctx)) @@ -1621,7 +1352,7 @@ static enum packet_return packet_read_register(struct gdb_context* gdbctx) assert(gdbctx->in_trap); reg = hex_to_int(gdbctx->in_packet, gdbctx->in_packet_len); - if (reg >= cpu_num_regs) + if (reg >= gdbctx->process->be_cpu->gdb_num_regs) { if (gdbctx->trace & GDBPXY_TRC_COMMAND_ERROR) fprintf(stderr, "Register out of bounds %x\n", reg); @@ -1634,9 +1365,9 @@ static enum packet_return packet_read_register(struct gdb_context* gdbctx) } if (gdbctx->trace & GDBPXY_TRC_COMMAND) { - if (cpu_register_map[reg].ctx_length <= sizeof(DWORD64)) + if (gdbctx->process->be_cpu->gdb_register_map[reg].ctx_length <= sizeof(DWORD64)) fprintf(stderr, "Read register %x => %08x%08x\n", reg, - (unsigned)(cpu_register(pctx, reg) >> 32), (unsigned)cpu_register(pctx, reg)); + (unsigned)(cpu_register(gdbctx, pctx, reg) >> 32), (unsigned)cpu_register(gdbctx, pctx, reg)); else fprintf(stderr, "Read register %x\n", reg); } @@ -1656,7 +1387,7 @@ static enum packet_return packet_write_register(struct gdb_context* gdbctx) assert(gdbctx->in_trap); reg = strtoul(gdbctx->in_packet, &ptr, 16); - if (ptr == NULL || reg >= cpu_num_regs || *ptr++ != '=') + if (ptr == NULL || reg >= gdbctx->process->be_cpu->gdb_num_regs || *ptr++ != '=') { if (gdbctx->trace & GDBPXY_TRC_COMMAND_ERROR) fprintf(stderr, "Invalid register index %s\n", gdbctx->in_packet); @@ -1677,7 +1408,7 @@ static enum packet_return packet_write_register(struct gdb_context* gdbctx) return packet_error; } - cpu_register_hex_from(pctx, reg, (const char**)&ptr); + cpu_register_hex_from(gdbctx, pctx, reg, (const char**)&ptr); if (pctx != &gdbctx->context && !gdbctx->process->be_cpu->set_context(gdbctx->other_thread->handle, pctx)) { -- 2.7.4

June 13, 2018

[PATCH 1/3] winedbg: Implement debugging 32-bit processes on x86_64.

by Zebediah Figura

Signed-off-by: Zebediah Figura <zfigura(a)codeweavers.com> --- programs/winedbg/be_i386.c | 22 ++++++++++++---------- programs/winedbg/db_disasm.c | 4 ++-- programs/winedbg/winedbg.c | 21 ++++++++------------- 3 files changed, 22 insertions(+), 25 deletions(-) diff --git a/programs/winedbg/be_i386.c b/programs/winedbg/be_i386.c index f46ebb1..4e279ec 100644 --- a/programs/winedbg/be_i386.c +++ b/programs/winedbg/be_i386.c @@ -21,7 +21,7 @@ #include "debugger.h" #include "wine/debug.h" -#ifdef __i386__ +#if defined(__i386__) || defined(__x86_64__) WINE_DEFAULT_DEBUG_CHANNEL(winedbg); @@ -33,6 +33,7 @@ extern void be_i386_disasm_one_insn(ADDRESS64* addr, int display); #define IS_VM86_MODE(ctx) (ctx->EFlags & V86_FLAG) +#ifndef __x86_64__ typedef struct DECLSPEC_ALIGN(16) _M128A { ULONGLONG Low; LONGLONG High; @@ -56,6 +57,7 @@ typedef struct _XMM_SAVE_AREA32 { M128A XmmRegisters[16]; /* 0a0 */ BYTE Reserved4[96]; /* 1a0 */ } XMM_SAVE_AREA32, *PXMM_SAVE_AREA32; +#endif static ADDRESS_MODE get_selector_type(HANDLE hThread, const WOW64_CONTEXT *ctx, WORD sel) { @@ -77,18 +79,18 @@ static void* be_i386_linearize(HANDLE hThread, const ADDRESS64* addr) switch (addr->Mode) { case AddrModeReal: - return (void*)((DWORD)(LOWORD(addr->Segment) << 4) + (DWORD)addr->Offset); + return (void*)((DWORD_PTR)(LOWORD(addr->Segment) << 4) + (DWORD_PTR)addr->Offset); case AddrMode1632: if (!(addr->Segment & 4) || ((addr->Segment >> 3) < 17)) - return (void*)(DWORD)addr->Offset; + return (void*)(DWORD_PTR)addr->Offset; /* fall through */ case AddrMode1616: if (!dbg_curr_process->process_io->get_selector(hThread, addr->Segment, &le)) return NULL; return (void*)((le.HighWord.Bits.BaseHi << 24) + (le.HighWord.Bits.BaseMid << 16) + le.BaseLow + - (DWORD)addr->Offset); + (DWORD_PTR)addr->Offset); case AddrModeFlat: - return (void*)(DWORD)addr->Offset; + return (void*)(DWORD_PTR)addr->Offset; } return NULL; } @@ -502,7 +504,7 @@ static BOOL be_i386_is_func_call(const void* insn, ADDRESS64* callee) if (!fetch_value((const char*)insn + 1, operand_size, &delta)) return FALSE; callee->Segment = dbg_context.x86.SegCs; - callee->Offset = (DWORD)insn + 1 + (operand_size / 8) + delta; + callee->Offset = (DWORD_PTR)insn + 1 + (operand_size / 8) + delta; return TRUE; case 0x9a: /* absolute far call */ @@ -589,11 +591,11 @@ static BOOL be_i386_is_func_call(const void* insn, ADDRESS64* callee) } if (((ch >> 3) & 0x07) == 0x03) /* LCALL */ { - if (!dbg_read_memory((const char*)dst + operand_size, &segment, sizeof(segment))) + if (!dbg_read_memory((const char*)(UINT_PTR)dst + operand_size, &segment, sizeof(segment))) return FALSE; } else segment = dbg_context.x86.SegCs; - if (!dbg_read_memory((const char*)dst, &delta, sizeof(delta))) + if (!dbg_read_memory((const char*)(UINT_PTR)dst, &delta, sizeof(delta))) return FALSE; callee->Mode = get_selector_type(dbg_curr_thread->handle, &dbg_context.x86, segment); @@ -643,7 +645,7 @@ static BOOL be_i386_is_jump(const void* insn, ADDRESS64* jumpee) if (!fetch_value((const char*)insn + 1, operand_size, &delta)) return FALSE; jumpee->Segment = dbg_context.x86.SegCs; - jumpee->Offset = (DWORD)insn + 1 + (operand_size / 8) + delta; + jumpee->Offset = (DWORD_PTR)insn + 1 + (operand_size / 8) + delta; return TRUE; default: WINE_FIXME("unknown %x\n", ch); return FALSE; } @@ -733,7 +735,7 @@ static BOOL be_i386_insert_Xpoint(HANDLE hProcess, const struct be_process_io* p bits = DR7_RW_WRITE; hw_bp: if ((reg = be_i386_get_unused_DR(ctx, &pr)) == -1) return FALSE; - *pr = (DWORD)addr; + *pr = (DWORD_PTR)addr; if (type != be_xpoint_watch_exec) switch (size) { case 4: bits |= DR7_LEN_4; break; diff --git a/programs/winedbg/db_disasm.c b/programs/winedbg/db_disasm.c index 9bdfb90..f57f009 100644 --- a/programs/winedbg/db_disasm.c +++ b/programs/winedbg/db_disasm.c @@ -64,7 +64,7 @@ #include "debugger.h" -#ifdef __i386__ +#if defined(__i386__) || defined(__x86_64__) /* * Switch to disassemble 16-bit code. @@ -1225,7 +1225,7 @@ static void db_print_address(const char *seg, int size, struct i_addr *addrp, in void* a2; dbg_printf("0x%x -> ", addrp->disp); - if (!dbg_read_memory((void*)addrp->disp, &a1, sizeof(a1))) { + if (!dbg_read_memory((void*)(INT_PTR)addrp->disp, &a1, sizeof(a1))) { dbg_printf("(invalid source)"); } else if (!dbg_read_memory(a1, &a2, sizeof(a2))) { dbg_printf("(invalid destination)"); diff --git a/programs/winedbg/winedbg.c b/programs/winedbg/winedbg.c index b5ffd76..1ac4187 100644 --- a/programs/winedbg/winedbg.c +++ b/programs/winedbg/winedbg.c @@ -284,6 +284,7 @@ extern struct backend_cpu be_i386; #elif defined(__powerpc__) extern struct backend_cpu be_ppc; #elif defined(__x86_64__) +extern struct backend_cpu be_i386; extern struct backend_cpu be_x86_64; #elif defined(__arm__) && !defined(__ARMEB__) extern struct backend_cpu be_arm; @@ -296,21 +297,13 @@ extern struct backend_cpu be_arm64; struct dbg_process* dbg_add_process(const struct be_process_io* pio, DWORD pid, HANDLE h) { struct dbg_process* p; + BOOL wow64; if ((p = dbg_get_process(pid))) - { - if (p->handle != 0) - { - WINE_ERR("Process (%04x) is already defined\n", pid); - } - else - { - p->handle = h; - p->process_io = pio; - p->imageName = NULL; - } return p; - } + + if (!h) + h = OpenProcess(PROCESS_ALL_ACCESS, FALSE, pid); if (!(p = HeapAlloc(GetProcessHeap(), 0, sizeof(struct dbg_process)))) return NULL; p->handle = h; @@ -333,12 +326,14 @@ struct dbg_process* dbg_add_process(const struct be_process_io* pio, DWORD pid, list_add_head(&dbg_process_list, &p->entry); + IsWow64Process(h, &wow64); + #ifdef __i386__ p->be_cpu = &be_i386; #elif defined(__powerpc__) p->be_cpu = &be_ppc; #elif defined(__x86_64__) - p->be_cpu = &be_x86_64; + p->be_cpu = wow64 ? &be_i386 : &be_x86_64; #elif defined(__arm__) && !defined(__ARMEB__) p->be_cpu = &be_arm; #elif defined(__aarch64__) && !defined(__AARCH64EB__) -- 2.7.4

June 13, 2018

Re: [PATCH 2/5] wined3d: Get rid of unnecessary enum wined3d_data_type values.

by Józef Kucia

On Wed, Jun 13, 2018 at 8:29 PM, Matteo Bruni <matteo.mystral(a)gmail.com> wrote: > 2018-06-13 10:37 GMT+02:00 Józef Kucia <joseph.kucia(a)gmail.com>: >> On Tue, Jun 12, 2018 at 6:41 PM, Matteo Bruni <mbruni(a)codeweavers.com> wrote: >>> diff --git a/dlls/wined3d/shader.c b/dlls/wined3d/shader.c >>> index 20d4f0773dd..a2c3f00559a 100644 >>> --- a/dlls/wined3d/shader.c >>> +++ b/dlls/wined3d/shader.c >>> @@ -2134,14 +2134,6 @@ static void shader_dump_decl_usage(struct wined3d_string_buffer *buffer, >>> shader_addline(buffer, " (uint)"); >>> break; >>> >>> - case WINED3D_DATA_UNORM: >>> - shader_addline(buffer, " (unorm)"); >>> - break; >>> - >>> - case WINED3D_DATA_SNORM: >>> - shader_addline(buffer, " (snorm)"); >>> - break; >>> - >> >> I don't like that it won't print the correct data type for >> declarations. For example, after the patch "dcl_uav_typed_texture2d >> (snorm) u0" is "dcl_uav_typed_texture2d (float) u0". > > Yeah, it's a downside of the "simplification". Would printing > something more generic like "(float / unorm / snorm)" work for you? > Any other idea? Perhaps it's possible to keep wined3d_data_type for wined3d_shader_semantic, or preserve the type information in wined3d_shader_semantic differently.

June 13, 2018

Re: [PATCH 6/6] dxgi: Get supported feature levels from wined3d caps.

by Józef Kucia

On Wed, Jun 13, 2018 at 8:31 PM, Matteo Bruni <matteo.mystral(a)gmail.com> wrote: > 2018-06-13 12:24 GMT+02:00 Józef Kucia <jkucia(a)codeweavers.com>: > >> On a side note, handling of feature levels 9_x appears to be incorrect >> (AFAIK feature levels 9_x always use SM2 shaders). > > True, but it's complicated... > Looking around https://msdn.microsoft.com/en-us/library/windows/desktop/ff476876(v=vs.85).… > you can see that feature level 9_3 e.g. mandates support for max > texture size >= 4096, which is only guaranteed by SM3-level hardware > (https://msdn.microsoft.com/en-us/library/windows/desktop/bb219845(v=vs.85).…) > IIRC there were a few of those, which is why I decided to require SM3 > for 9_3 even though the shaders themselves are going to be SM2 only. > Right, feature level checks should not be based solely on shader model versions.

June 13, 2018

[PATCH] winex11.drv: Refcount the vulkan surface window

by Andrew Eikum

Signed-off-by: Andrew Eikum <aeikum(a)codeweavers.com> --- Modeled after the opengl refcounting. This prevents us getting X errors from using client_window after it's been destroyed. dlls/winex11.drv/vulkan.c | 61 ++++++++++++++++++++++++++++++++++----- dlls/winex11.drv/window.c | 1 + dlls/winex11.drv/x11drv.h | 1 + 3 files changed, 55 insertions(+), 8 deletions(-) diff --git a/dlls/winex11.drv/vulkan.c b/dlls/winex11.drv/vulkan.c index 7e252326b5..7eca31bdea 100644 --- a/dlls/winex11.drv/vulkan.c +++ b/dlls/winex11.drv/vulkan.c @@ -41,11 +41,23 @@ WINE_DEFAULT_DEBUG_CHANNEL(vulkan); #ifdef SONAME_LIBVULKAN +static CRITICAL_SECTION context_section; +static CRITICAL_SECTION_DEBUG critsect_debug = +{ + 0, 0, &context_section, + { &critsect_debug.ProcessLocksList, &critsect_debug.ProcessLocksList }, + 0, 0, { (DWORD_PTR)(__FILE__ ": context_section") } +}; +static CRITICAL_SECTION context_section = { &critsect_debug, -1, 0, 0, 0, 0 }; + +static XContext vulkan_hwnd_context; + typedef VkFlags VkXlibSurfaceCreateFlagsKHR; #define VK_STRUCTURE_TYPE_XLIB_SURFACE_CREATE_INFO_KHR 1000004000 struct wine_vk_surface { + LONG ref; Window window; VkSurfaceKHR surface; /* native surface */ }; @@ -115,6 +127,8 @@ static BOOL WINAPI wine_vk_init(INIT_ONCE *once, void *param, void **context) LOAD_FUNCPTR(vkQueuePresentKHR) #undef LOAD_FUNCPTR + vulkan_hwnd_context = XUniqueContext(); + return TRUE; fail: @@ -171,20 +185,35 @@ static VkResult wine_vk_instance_convert_create_info(const VkInstanceCreateInfo return VK_SUCCESS; } -static void wine_vk_surface_destroy(VkInstance instance, struct wine_vk_surface *surface) +static struct wine_vk_surface *wine_vk_surface_grab(struct wine_vk_surface *surface) { - if (!surface) + InterlockedIncrement(&surface->ref); + return surface; +} + +static void wine_vk_surface_release(struct wine_vk_surface *surface) +{ + if (InterlockedDecrement(&surface->ref)) return; - /* vkDestroySurfaceKHR must handle VK_NULL_HANDLE (0) for surface. */ - pvkDestroySurfaceKHR(instance, surface->surface, NULL /* allocator */); - if (surface->window) XDestroyWindow(gdi_display, surface->window); heap_free(surface); } +void wine_vk_surface_destroy(HWND hwnd) +{ + struct wine_vk_surface *surface; + EnterCriticalSection(&context_section); + if (!XFindContext(gdi_display, (XID)hwnd, vulkan_hwnd_context, (char **)&surface)) + { + wine_vk_surface_release(surface); + } + XDeleteContext(gdi_display, (XID)hwnd, vulkan_hwnd_context); + LeaveCriticalSection(&context_section); +} + static VkResult X11DRV_vkAcquireNextImageKHR(VkDevice device, VkSwapchainKHR swapchain, uint64_t timeout, VkSemaphore semaphore, VkFence fence, uint32_t *index) { @@ -245,7 +274,7 @@ static VkResult X11DRV_vkCreateWin32SurfaceKHR(VkInstance instance, { VkResult res; VkXlibSurfaceCreateInfoKHR create_info_host; - struct wine_vk_surface *x11_surface; + struct wine_vk_surface *x11_surface, *prev; TRACE("%p %p %p %p\n", instance, create_info, allocator, surface); @@ -263,6 +292,8 @@ static VkResult X11DRV_vkCreateWin32SurfaceKHR(VkInstance instance, if (!x11_surface) return VK_ERROR_OUT_OF_HOST_MEMORY; + x11_surface->ref = 1; + x11_surface->window = create_client_window(create_info->hwnd, &default_visual); if (!x11_surface->window) { @@ -286,13 +317,21 @@ static VkResult X11DRV_vkCreateWin32SurfaceKHR(VkInstance instance, goto err; } + EnterCriticalSection(&context_section); + if (!XFindContext(gdi_display, (XID)create_info->hwnd, vulkan_hwnd_context, (char **)&prev)) + { + wine_vk_surface_release(prev); + } + XSaveContext(gdi_display, (XID)create_info->hwnd, vulkan_hwnd_context, (char *)wine_vk_surface_grab(x11_surface)); + LeaveCriticalSection(&context_section); + *surface = (uintptr_t)x11_surface; TRACE("Created surface=0x%s\n", wine_dbgstr_longlong(*surface)); return VK_SUCCESS; err: - wine_vk_surface_destroy(instance, x11_surface); + wine_vk_surface_release(x11_surface); return res; } @@ -316,7 +355,13 @@ static void X11DRV_vkDestroySurfaceKHR(VkInstance instance, VkSurfaceKHR surface if (allocator) FIXME("Support for allocation callbacks not implemented yet\n"); - wine_vk_surface_destroy(instance, x11_surface); + /* vkDestroySurfaceKHR must handle VK_NULL_HANDLE (0) for surface. */ + if (x11_surface) + { + pvkDestroySurfaceKHR(instance, x11_surface->surface, NULL /* allocator */); + + wine_vk_surface_release(x11_surface); + } } static void X11DRV_vkDestroySwapchainKHR(VkDevice device, VkSwapchainKHR swapchain, diff --git a/dlls/winex11.drv/window.c b/dlls/winex11.drv/window.c index 5fecd9a17b..a848dc61e1 100644 --- a/dlls/winex11.drv/window.c +++ b/dlls/winex11.drv/window.c @@ -1719,6 +1719,7 @@ void CDECL X11DRV_DestroyWindow( HWND hwnd ) release_win_data( data ); HeapFree( GetProcessHeap(), 0, data ); destroy_gl_drawable( hwnd ); + wine_vk_surface_destroy( hwnd ); } diff --git a/dlls/winex11.drv/x11drv.h b/dlls/winex11.drv/x11drv.h index 85a05a904a..939c2e7f3b 100644 --- a/dlls/winex11.drv/x11drv.h +++ b/dlls/winex11.drv/x11drv.h @@ -586,6 +586,7 @@ extern XIC X11DRV_get_ic( HWND hwnd ) DECLSPEC_HIDDEN; extern void sync_gl_drawable( HWND hwnd ) DECLSPEC_HIDDEN; extern void set_gl_drawable_parent( HWND hwnd, HWND parent ) DECLSPEC_HIDDEN; extern void destroy_gl_drawable( HWND hwnd ) DECLSPEC_HIDDEN; +extern void wine_vk_surface_destroy( HWND hwnd ) DECLSPEC_HIDDEN; extern void wait_for_withdrawn_state( HWND hwnd, BOOL set ) DECLSPEC_HIDDEN; extern Window init_clip_window(void) DECLSPEC_HIDDEN; -- 2.17.1

June 13, 2018

Re: [PATCH 1/2] winmm/tests: Avoid and ARRAY_SIZE-like macro

by Marvin

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=39112 Your paranoid android. === w8 (32 bit mmio) === TestBot process got stuck or died unexpectedly The previous 1 run(s) terminated abnormally

June 13, 2018

Re: [PATCH 1/2] dsound/tests: Avoid ARRAY_SIZE-like macros

by Michael Stefaniuc

On 06/13/2018 08:54 PM, Marvin wrote: > While running your changed tests on Windows, I think I found new failures. > Being a bot and all I'm not very good at pattern recognition, so I might be > wrong, but could you please double-check? > Full results can be found at > https://testbot.winehq.org/JobDetails.pl?Key=39109 > > Your paranoid android. > > > === w8 (32 bit ds3d8) === > TestBot process got stuck or died unexpectedly > The previous 1 run(s) terminated abnormally Testbot issue, looks like the start time is missing: 0e90:ds3d8: 757 tests executed (0 marked as todo, 0 failures), 0 skipped. dsound:ds3d8:0e90 done (0) in 0s bye michael

June 13, 2018

Re: [PATCH 1/2] dsound/tests: Avoid ARRAY_SIZE-like macros

by Marvin

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=39109 Your paranoid android. === w8 (32 bit ds3d8) === TestBot process got stuck or died unexpectedly The previous 1 run(s) terminated abnormally

June 13, 2018

Re: [PATCH 6/6] dxgi: Get supported feature levels from wined3d caps.

by Matteo Bruni

2018-06-13 12:24 GMT+02:00 Józef Kucia <jkucia(a)codeweavers.com>: > On a side note, handling of feature levels 9_x appears to be incorrect > (AFAIK feature levels 9_x always use SM2 shaders). True, but it's complicated... Looking around https://msdn.microsoft.com/en-us/library/windows/desktop/ff476876(v=vs.85).… you can see that feature level 9_3 e.g. mandates support for max texture size >= 4096, which is only guaranteed by SM3-level hardware (https://msdn.microsoft.com/en-us/library/windows/desktop/bb219845(v=vs.85).…) IIRC there were a few of those, which is why I decided to require SM3 for 9_3 even though the shaders themselves are going to be SM2 only.

June 13, 2018