x64 Calling convetion(ABI) issues.
in [Device Drivers]
|
From: Pavel A. on 16 Aug 2005 07:00 So , this is the purpose of the HCT calling convention test? Check that exceptions will work properly? --PA "Maxim S. Shatskih" wrote: > > of AMD64, that is radically different, and that requires the > > compiler/assembler > > to properly describe the code regions > > IIRC "gcc" uses the same on x86 for C++ exceptions. This results in eliminating > nearly any extra machine code use to enter and leave the exception frames. > > -- > Maxim Shatskih, Windows DDK MVP > StorageCraft Corporation > maxim(a)storagecraft.com > http://www.storagecraft.com > > >
From: Ivan Brugiolo [MSFT] on 16 Aug 2005 12:08 Essentially, yes. To make sure that an exception can cross the driver's code without bug-checking the machine. -- This posting is provided "AS IS" with no warranties, and confers no rights. Use of any included script samples are subject to the terms specified at http://www.microsoft.com/info/cpyright.htm "Pavel A." <pavel_a(a)NOwritemeNO.com> wrote in message news:A4BC21D3-77AD-4E18-A117-61A3B6E05EB0(a)microsoft.com... > So , this is the purpose of the HCT calling convention test? > Check that exceptions will work properly? > > --PA > > > "Maxim S. Shatskih" wrote: >> > of AMD64, that is radically different, and that requires the >> > compiler/assembler >> > to properly describe the code regions >> >> IIRC "gcc" uses the same on x86 for C++ exceptions. This results in >> eliminating >> nearly any extra machine code use to enter and leave the exception >> frames. >> >> -- >> Maxim Shatskih, Windows DDK MVP >> StorageCraft Corporation >> maxim(a)storagecraft.com >> http://www.storagecraft.com >> >> >>
From: Tonald on 17 Aug 2005 21:37 The error is like this: " <Assert File="e:\amd64\test.asm" Line="28" UserText="Error 103: The stack must always be 16-byte aligned when a call instruction is executed. Misaligned stack with relative depth FFFFFF08 at instruction: @ call Do_FIR" CA="3418789856" LA="3419016946">" The function code attached below: What My finding is that, the test treat the stack position where call Do_FIR is a negative position. If i let the stack position be add rsp,28h, the test will have no such error, but of course, add rsp is fatal error. ..DATA ADDR1 QWORD 0 ADDR2 QWORD 0 ADDR3 QWORD 0 ADDR4 QWORD 0 ..CODE PUBLIC TEST PUBLIC Do_FIR TEST proc FRAME sub rsp 98h .allocstack 98h .endprolog ... .... ..... mov r9,ADD4 mov r8,ADD3 mov rdx,ADD2 mov rcx,ADD1 mov QWORD PTR [rsp+18h],r9 mov QWORD PTR [rsp+10h],r8 mov QWORD PTR [rsp+8h],rdx mov QWORD PTR [rsp],rcx call Do_FIR ... .... ..... add rsp,98h ret TEST endp END "Ivan Brugiolo [MSFT]" wrote: > Essentially, yes. > To make sure that an exception can cross the driver's code > without bug-checking the machine. > > -- > This posting is provided "AS IS" with no warranties, and confers no rights. > Use of any included script samples are subject to the terms specified at > http://www.microsoft.com/info/cpyright.htm > > > "Pavel A." <pavel_a(a)NOwritemeNO.com> wrote in message > news:A4BC21D3-77AD-4E18-A117-61A3B6E05EB0(a)microsoft.com... > > So , this is the purpose of the HCT calling convention test? > > Check that exceptions will work properly? > > > > --PA > > > > > > "Maxim S. Shatskih" wrote: > >> > of AMD64, that is radically different, and that requires the > >> > compiler/assembler > >> > to properly describe the code regions > >> > >> IIRC "gcc" uses the same on x86 for C++ exceptions. This results in > >> eliminating > >> nearly any extra machine code use to enter and leave the exception > >> frames. > >> > >> -- > >> Maxim Shatskih, Windows DDK MVP > >> StorageCraft Corporation > >> maxim(a)storagecraft.com > >> http://www.storagecraft.com > >> > >> > >> > > >
From: Ivan Brugiolo [MSFT] on 26 Aug 2005 13:10 I have replicated the problem, and the owner of abiprov.exe will try to address the issue in the tool. -- This posting is provided "AS IS" with no warranties, and confers no rights. Use of any included script samples are subject to the terms specified at http://www.microsoft.com/info/cpyright.htm "Tonald" <Tonald(a)discussions.microsoft.com> wrote in message news:67849938-58D8-42EC-9BEF-7A5EB7945CE1(a)microsoft.com... > The error is like this: > " <Assert File="e:\amd64\test.asm" Line="28" UserText="Error 103: The > stack > must always be 16-byte aligned when a call instruction is executed. > Misaligned stack with relative depth FFFFFF08 at instruction: @ call > Do_FIR" > CA="3418789856" LA="3419016946">" > > The function code attached below: > What My finding is that, the test treat the stack position where call > Do_FIR > is a negative position. If i let the stack position be add rsp,28h, the > test > will have no such error, but of course, add rsp is fatal error. > > > > > .DATA > ADDR1 QWORD 0 > ADDR2 QWORD 0 > ADDR3 QWORD 0 > ADDR4 QWORD 0 > .CODE > PUBLIC TEST > PUBLIC Do_FIR > > TEST proc FRAME > > sub rsp 98h > .allocstack 98h > .endprolog > > > > ... > .... > ..... > mov r9,ADD4 > mov r8,ADD3 > mov rdx,ADD2 > mov rcx,ADD1 > > mov QWORD PTR [rsp+18h],r9 > mov QWORD PTR [rsp+10h],r8 > mov QWORD PTR [rsp+8h],rdx > mov QWORD PTR [rsp],rcx > > call Do_FIR > > > ... > .... > ..... > > add rsp,98h > ret > TEST endp > > END > > > > "Ivan Brugiolo [MSFT]" wrote: > >> Essentially, yes. >> To make sure that an exception can cross the driver's code >> without bug-checking the machine. >> >> -- >> This posting is provided "AS IS" with no warranties, and confers no >> rights. >> Use of any included script samples are subject to the terms specified at >> http://www.microsoft.com/info/cpyright.htm >> >> >> "Pavel A." <pavel_a(a)NOwritemeNO.com> wrote in message >> news:A4BC21D3-77AD-4E18-A117-61A3B6E05EB0(a)microsoft.com... >> > So , this is the purpose of the HCT calling convention test? >> > Check that exceptions will work properly? >> > >> > --PA >> > >> > >> > "Maxim S. Shatskih" wrote: >> >> > of AMD64, that is radically different, and that requires the >> >> > compiler/assembler >> >> > to properly describe the code regions >> >> >> >> IIRC "gcc" uses the same on x86 for C++ exceptions. This results in >> >> eliminating >> >> nearly any extra machine code use to enter and leave the exception >> >> frames. >> >> >> >> -- >> >> Maxim Shatskih, Windows DDK MVP >> >> StorageCraft Corporation >> >> maxim(a)storagecraft.com >> >> http://www.storagecraft.com >> >> >> >> >> >> >> >> >>
First
|
Prev
|
Pages: 1 2 Prev: USB device access via DeviceIoControl and WriteFile / ReadFile Next: PcCardConfig M Option |