pc-bios/s390-ccw: Fixes and improvements for start.S
v2 cover pc-bios/s390-ccw: Fixes and improvements for start.S
Main motivation of this series was a bug that showed up when compiling with Clang 16 and binutils 2.40 (which has been reported in Fedora ELN, see
https://bugzilla.redhat.com/show_bug.cgi?id=2216662). This is fixed in the fourth patch. I checked with "objdump" that the change is fine, indeed.
While working on this issue, I came accross some other issues which I address in the first three patches:
- Indentation is a mixture between tabs and spaces in start.S (patch 1)
- We do not set up a stack frame for the main() function, which could cause memory corruption (patch 2)
- The stack is declared in multiple places, though it's only needed in start.S (patch 3)
v2:
- Use ".space" instead of ".lcomm" in the third patch to make sure that the alignment is really taken into consideration (thanks Richard)
- Alignment of 8 should be enough in the third patch (thank Christian)
- Added Reviewed-bys from v1
Thomas Huth (4):
pc-bios/s390-ccw: Fix indentation in start.S pc-bios/s390-ccw: Provide space for initial stack frame in start.S pc-bios/s390-ccw: Move the stack array into start.S pc-bios/s390-ccw: Don't use __bss_start with the "larl" instruction
pc-bios/s390-ccw/s390-ccw.h | 1 - pc-bios/s390-ccw/main.c | 1 - pc-bios/s390-ccw/netmain.c | 1 - pc-bios/s390-ccw/start.S | 144 +++++++++++++++++++----------------- 4 files changed, 76 insertions(+), 71 deletions(-)
Signature · 1 line
2.39.3
PATCH v2 1/4 pc-bios/s390-ccw: Fix indentation in start.S
start.S is currently indented with a mixture of spaces and tabs, which is quite ugly. QEMU coding style says indentation should be 4 spaces, and this is also what we are using in the assembler files in the tests/tcg/s390x/ folder already, so let's adjust start.S accordingly.
--- pc-bios/s390-ccw/start.S | 136 +++++++++++++++++++-------------------- 1 file changed, 68 insertions(+), 68 deletions(-)
pc-bios/s390-ccw/start.S +68 -68
@@ -10,37 +10,37 @@ * directory. */ - .globl _start + .globl _start _start: - larl %r15, stack + 0x8000 /* Set up stack */ + larl %r15,stack + 0x8000 /* Set up stack */ - /* clear bss */ - larl %r2, __bss_start - larl %r3, _end - slgr %r3, %r2 /* get sizeof bss */ - ltgr %r3,%r3 /* bss empty? */ - jz done - aghi %r3,-1 - srlg %r4,%r3,8 /* how many 256 byte chunks? */ - ltgr %r4,%r4 - lgr %r1,%r2 - jz remainder + /* clear bss */ + larl %r2,__bss_start + larl %r3,_end + slgr %r3,%r2 /* get sizeof bss */ + ltgr %r3,%r3 /* bss empty? */ + jz done + aghi %r3,-1 + srlg %r4,%r3,8 /* how many 256 byte chunks? */ + ltgr %r4,%r4 + lgr %r1,%r2 + jz remainder loop: - xc 0(256,%r1),0(%r1) - la %r1,256(%r1) - brctg %r4,loop + xc 0(256,%r1),0(%r1) + la %r1,256(%r1) + brctg %r4,loop remainder: - larl %r2,memsetxc - ex %r3,0(%r2) + larl %r2,memsetxc + ex %r3,0(%r2) done: - /* set up a pgm exception disabled wait psw */ - larl %r2, disabled_wait_psw - mvc 0x01d0(16), 0(%r2) - j main /* And call C */ + /* set up a pgm exception disabled wait psw */ + larl %r2,disabled_wait_psw + mvc 0x01d0(16),0(%r2) + j main /* And call C */ memsetxc: - xc 0(1,%r1),0(%r1) + xc 0(1,%r1),0(%r1) /*
@@ -48,11 +48,11 @@ memsetxc: * * stops the current guest cpu. */ - .globl disabled_wait + .globl disabled_wait disabled_wait: - larl %r1,disabled_wait_psw - lpswe 0(%r1) -1: j 1b + larl %r1,disabled_wait_psw + lpswe 0(%r1) +1: j 1b /*
@@ -60,61 +60,61 @@ disabled_wait: * * eats one sclp interrupt */ - .globl consume_sclp_int + .globl consume_sclp_int consume_sclp_int: - /* enable service interrupts in cr0 */ - stctg %c0,%c0,0(%r15) - oi 6(%r15),0x2 - lctlg %c0,%c0,0(%r15) - /* prepare external call handler */ - larl %r1, external_new_code - stg %r1, 0x1b8 - larl %r1, external_new_mask - mvc 0x1b0(8),0(%r1) - /* load enabled wait PSW */ - larl %r1, enabled_wait_psw - lpswe 0(%r1) + /* enable service interrupts in cr0 */ + stctg %c0,%c0,0(%r15) + oi 6(%r15),0x2 + lctlg %c0,%c0,0(%r15) + /* prepare external call handler */ + larl %r1,external_new_code + stg %r1,0x1b8 + larl %r1,external_new_mask + mvc 0x1b0(8),0(%r1) + /* load enabled wait PSW */ + larl %r1,enabled_wait_psw + lpswe 0(%r1) /* * void consume_io_int(void) * * eats one I/O interrupt */ - .globl consume_io_int + .globl consume_io_int consume_io_int: - /* enable I/O interrupts in cr6 */ - stctg %c6,%c6,0(%r15) - oi 4(%r15), 0xff - lctlg %c6,%c6,0(%r15) - /* prepare i/o call handler */ - larl %r1, io_new_code - stg %r1, 0x1f8 - larl %r1, io_new_mask - mvc 0x1f0(8),0(%r1) - /* load enabled wait PSW */ - larl %r1, enabled_wait_psw - lpswe 0(%r1) + /* enable I/O interrupts in cr6 */ + stctg %c6,%c6,0(%r15) + oi 4(%r15), 0xff + lctlg %c6,%c6,0(%r15) + /* prepare i/o call handler */ + larl %r1,io_new_code + stg %r1,0x1f8 + larl %r1,io_new_mask + mvc 0x1f0(8),0(%r1) + /* load enabled wait PSW */ + larl %r1,enabled_wait_psw + lpswe 0(%r1) external_new_code: - /* disable service interrupts in cr0 */ - stctg %c0,%c0,0(%r15) - ni 6(%r15),0xfd - lctlg %c0,%c0,0(%r15) - br %r14 + /* disable service interrupts in cr0 */ + stctg %c0,%c0,0(%r15) + ni 6(%r15),0xfd + lctlg %c0,%c0,0(%r15) + br %r14 io_new_code: - /* disable I/O interrupts in cr6 */ - stctg %c6,%c6,0(%r15) - ni 4(%r15), 0x00 - lctlg %c6,%c6,0(%r15) - br %r14 + /* disable I/O interrupts in cr6 */ + stctg %c6,%c6,0(%r15) + ni 4(%r15),0x00 + lctlg %c6,%c6,0(%r15) + br %r14 - .align 8 + .align 8 disabled_wait_psw: - .quad 0x0002000180000000,0x0000000000000000 + .quad 0x0002000180000000,0x0000000000000000 enabled_wait_psw: - .quad 0x0302000180000000,0x0000000000000000 + .quad 0x0302000180000000,0x0000000000000000 external_new_mask: - .quad 0x0000000180000000 + .quad 0x0000000180000000 io_new_mask: - .quad 0x0000000180000000 + .quad 0x0000000180000000
Signature · 1 line
2.39.3
Re: [PATCH v2 1/4] pc-bios/s390-ccw: Fix indentation in start.S
Quoted 10 lines · Thomas Huth
start.Sis currently indented with a mixture of spaces and tabs, which is quite ugly. QEMU coding style says indentation should be 4 spaces, and this is also what we are using in the assembler files in the tests/tcg/s390x/ folder already, so let's adjuststart.Saccordingly.Reviewed-by Cédric Le Goater <[email protected]>Signed-off-by Thomas Huth <[email protected]>--- pc-bios/s390-ccw/start.S | 136 +++++++++++++++++++-------------------- 1 file changed, 68 insertions(+), 68 deletions(-)
Re: [PATCH v2 1/4] pc-bios/s390-ccw: Fix indentation in start.S
In reply to Thomas HuthQuoted 7 lines · On Tue, 27 Jun 2023 09:47:00 +0200 Thomas Huth <[email protected]> wrote:
start.Sis currently indented with a mixture of spaces and tabs, which is quite ugly. QEMU coding style says indentation should be 4 spaces, and this is also what we are using in the assembler files in the tests/tcg/s390x/ folder already, so let's adjuststart.Saccordingly.Reviewed-by Cédric Le Goater <[email protected]>Signed-off-by Thomas Huth <[email protected]>
Quoted 188 lines
--- pc-bios/s390-ccw/start.S | 136 +++++++++++++++++++-------------------- 1 file changed, 68 insertions(+), 68 deletions(-)pc-bios/s390-ccw/start.S +68
-68@@ -10,37 +10,37 @@ * directory. */ - .globl _start + .globl _start _start: - larl %r15, stack + 0x8000 /* Set up stack */ + larl %r15,stack + 0x8000 /* Set up stack */ - /* clear bss */ - larl %r2, __bss_start - larl %r3, _end - slgr %r3, %r2 /* get sizeof bss */ - ltgr %r3,%r3 /* bss empty? */ - jz done - aghi %r3,-1 - srlg %r4,%r3,8 /* how many 256 byte chunks? */ - ltgr %r4,%r4 - lgr %r1,%r2 - jz remainder + /* clear bss */ + larl %r2,__bss_start + larl %r3,_end + slgr %r3,%r2 /* get sizeof bss */ + ltgr %r3,%r3 /* bss empty? */ + jz done + aghi %r3,-1 + srlg %r4,%r3,8 /* how many 256 byte chunks? */ + ltgr %r4,%r4 + lgr %r1,%r2 + jz remainder loop: - xc 0(256,%r1),0(%r1) - la %r1,256(%r1) - brctg %r4,loop + xc 0(256,%r1),0(%r1) + la %r1,256(%r1) + brctg %r4,loop remainder: - larl %r2,memsetxc - ex %r3,0(%r2) + larl %r2,memsetxc + ex %r3,0(%r2) done: - /* set up a pgm exception disabled wait psw */ - larl %r2, disabled_wait_psw - mvc 0x01d0(16), 0(%r2) - j main /* And call C */ + /* set up a pgm exception disabled wait psw */ + larl %r2,disabled_wait_psw + mvc 0x01d0(16),0(%r2) + j main /* And call C */ memsetxc: - xc 0(1,%r1),0(%r1) + xc 0(1,%r1),0(%r1) /*@@ -48,11 +48,11 @@ memsetxc: * * stops the current guest cpu. */ - .globl disabled_wait + .globl disabled_wait disabled_wait: - larl %r1,disabled_wait_psw - lpswe 0(%r1) -1: j 1b + larl %r1,disabled_wait_psw + lpswe 0(%r1) +1: j 1b /*@@ -60,61 +60,61 @@ disabled_wait: * * eats one sclp interrupt */ - .globl consume_sclp_int + .globl consume_sclp_int consume_sclp_int: - /* enable service interrupts in cr0 */ - stctg %c0,%c0,0(%r15) - oi 6(%r15),0x2 - lctlg %c0,%c0,0(%r15) - /* prepare external call handler */ - larl %r1, external_new_code - stg %r1, 0x1b8 - larl %r1, external_new_mask - mvc 0x1b0(8),0(%r1) - /* load enabled wait PSW */ - larl %r1, enabled_wait_psw - lpswe 0(%r1) + /* enable service interrupts in cr0 */ + stctg %c0,%c0,0(%r15) + oi 6(%r15),0x2 + lctlg %c0,%c0,0(%r15) + /* prepare external call handler */ + larl %r1,external_new_code + stg %r1,0x1b8 + larl %r1,external_new_mask + mvc 0x1b0(8),0(%r1) + /* load enabled wait PSW */ + larl %r1,enabled_wait_psw + lpswe 0(%r1) /* * void consume_io_int(void) * * eats one I/O interrupt */ - .globl consume_io_int + .globl consume_io_int consume_io_int: - /* enable I/O interrupts in cr6 */ - stctg %c6,%c6,0(%r15) - oi 4(%r15), 0xff - lctlg %c6,%c6,0(%r15) - /* prepare i/o call handler */ - larl %r1, io_new_code - stg %r1, 0x1f8 - larl %r1, io_new_mask - mvc 0x1f0(8),0(%r1) - /* load enabled wait PSW */ - larl %r1, enabled_wait_psw - lpswe 0(%r1) + /* enable I/O interrupts in cr6 */ + stctg %c6,%c6,0(%r15) + oi 4(%r15), 0xff + lctlg %c6,%c6,0(%r15) + /* prepare i/o call handler */ + larl %r1,io_new_code + stg %r1,0x1f8 + larl %r1,io_new_mask + mvc 0x1f0(8),0(%r1) + /* load enabled wait PSW */ + larl %r1,enabled_wait_psw + lpswe 0(%r1) external_new_code: - /* disable service interrupts in cr0 */ - stctg %c0,%c0,0(%r15) - ni 6(%r15),0xfd - lctlg %c0,%c0,0(%r15) - br %r14 + /* disable service interrupts in cr0 */ + stctg %c0,%c0,0(%r15) + ni 6(%r15),0xfd + lctlg %c0,%c0,0(%r15) + br %r14 io_new_code: - /* disable I/O interrupts in cr6 */ - stctg %c6,%c6,0(%r15) - ni 4(%r15), 0x00 - lctlg %c6,%c6,0(%r15) - br %r14 + /* disable I/O interrupts in cr6 */ + stctg %c6,%c6,0(%r15) + ni 4(%r15),0x00 + lctlg %c6,%c6,0(%r15) + br %r14 - .align 8 + .align 8 disabled_wait_psw: - .quad 0x0002000180000000,0x0000000000000000 + .quad 0x0002000180000000,0x0000000000000000 enabled_wait_psw: - .quad 0x0302000180000000,0x0000000000000000 + .quad 0x0302000180000000,0x0000000000000000 external_new_mask: - .quad 0x0000000180000000 + .quad 0x0000000180000000 io_new_mask: - .quad 0x0000000180000000 + .quad 0x0000000180000000
Re: [PATCH v2 1/4] pc-bios/s390-ccw: Fix indentation in start.S
In reply to Thomas HuthQuoted 6 lines · Thomas Huth
start.Sis currently indented with a mixture of spaces and tabs, which
is quite ugly. QEMU coding style says indentation should be 4 spaces, and this is also what we are using in the assembler files in the tests/tcg/s390x/ folder already, so let's adjuststart.Saccordingly.
Quoted 6 lines
Reviewed-by Cédric Le Goater <[email protected]>Signed-off-by Thomas Huth <[email protected]>---
pc-bios/s390-ccw/start.S | 136 +++++++++++++++++++------------------ --1 file changed, 68 insertions(+), 68 deletions(-)
Oh, that looks nice. Thank you.
PATCH v2 2/4 pc-bios/s390-ccw: Provide space for initial stack frame in start.S
In reply to Thomas HuthProviding the space of a stack frame is the duty of the caller, so we should reserve 160 bytes before jumping into the main function. Otherwise the main() function might write past the stack array.
While we're at it, add a proper STACK_SIZE macro for the stack size instead of using magic numbers (this is also required for the following patch).
--- pc-bios/s390-ccw/start.S | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
pc-bios/s390-ccw/start.S +3 -1
@@ -10,10 +10,12 @@ * directory. */ +#define STACK_SIZE 0x8000 + .globl _start _start: - larl %r15,stack + 0x8000 /* Set up stack */ + larl %r15,stack + STACK_SIZE - 160 /* Set up stack */ /* clear bss */ larl %r2,__bss_start
Signature · 1 line
2.39.3
Re: [PATCH v2 2/4] pc-bios/s390-ccw: Provide space for initial stack frame in start.S
Quoted 20 lines · Thomas Huth
Providing the space of a stack frame is the duty of the caller, so we should reserve 160 bytes before jumping into the main function. Otherwise the
main()function might write past the stack array.While we're at it, add a proper
STACK_SIZEmacro for the stack size instead of using magic numbers (this is also required for the following patch).Reviewed-by Christian Borntraeger <[email protected]>Reviewed-by Cédric Le Goater <[email protected]>Signed-off-by Thomas Huth <[email protected]>--- pc-bios/s390-ccw/start.S | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)pc-bios/s390-ccw/start.S +0
-0
pc-bios/s390-ccw/start.S line 18
Quoted 10 lines
pc-bios/s390-ccw/start.S +3
-1@@ -10,10 +10,12 @@ * directory. */ +#define STACK_SIZE 0x8000 + .globl _start _start: - larl %r15,stack + 0x8000 /* Set up stack */ + larl %r15,stack + STACK_SIZE - 160 /* Set up stack */
^^^
You can also add a macro for this
- e.g. STACK_FRAME_SIZE.
Besides that,
Quoted 6 lines
pc-bios/s390-ccw/start.S +0
-0/* clear bss */ larl %r2,__bss_startSignature · 1 line
2.39.3
Signature · 8 lines
Kind regards / Beste Grüße Marc Hartmayer IBM Deutschland Research & Development GmbH Vorsitzender des Aufsichtsrats: Gregor Pillen Geschäftsführung: David Faller Sitz der Gesellschaft: Böblingen Registergericht: Amtsgericht Stuttgart, HRB 243294
Philippe Mathieu-Daudé
Reviewed-byRe: [PATCH v2 2/4] pc-bios/s390-ccw: Provide space for initial stack frame in start.S
Quoted 23 lines · Marc Hartmayer
Quoted 21 lines · Thomas Huth
Providing the space of a stack frame is the duty of the caller, so we should reserve 160 bytes before jumping into the main function. Otherwise the
main()function might write past the stack array.While we're at it, add a proper
STACK_SIZEmacro for the stack size instead of using magic numbers (this is also required for the following patch).Reviewed-by Christian Borntraeger <[email protected]>Reviewed-by Cédric Le Goater <[email protected]>Signed-off-by Thomas Huth <[email protected]>--- pc-bios/s390-ccw/start.S | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)pc-bios/s390-ccw/start.S +2
-0+#define STACK_SIZE 0x8000 + .globl _start _start:
Quoted 5 lines
pc-bios/s390-ccw/start.S line 18
Quoted 2 lines
pc-bios/s390-ccw/start.S +1
-1- larl %r15,stack + 0x8000 /* Set up stack */ + larl %r15,stack + STACK_SIZE - 160 /* Set up stack */^^^
You can also add a macro for this
- e.g.STACK_FRAME_SIZE.
Yes please :) No need to respin.
Quoted 2 lines
Besides that,
Reviewed-by Marc Hartmayer <[email protected]>
=
Eric Farman
Reviewed-byRe: [PATCH v2 2/4] pc-bios/s390-ccw: Provide space for initial stack frame in start.S
Quoted 34 lines · Philippe Mathieu-Daudé
Quoted 21 lines · Marc Hartmayer
Quoted 17 lines · Thomas Huth
Providing the space of a stack frame is the duty of the caller, so we should reserve 160 bytes before jumping into the main function.
Otherwise themain()function might write past the stack array.While we're at it, add a proper
STACK_SIZEmacro for the stack size
instead of using magic numbers (this is also required for the following
patch).Reviewed-by Christian Borntraeger <[email protected]>Reviewed-by Cédric Le Goater <[email protected]>Signed-off-by Thomas Huth <[email protected]>---
pc-bios/s390-ccw/start.S| 4 +++-1 file changed, 3 insertions(+), 1 deletion(-)Quoted 11 lines
Quoted 7 lines
pc-bios/s390-ccw/start.S +2
-0+#define STACK_SIZE 0x8000 +.globl
_start
_start:
- larl %r15,stack + 0x8000 /* Set up stack */ + larl %r15,stack +STACK_SIZE- 160 /* Set up stack */^^^
You can also add a macro for this
- e.g.STACK_FRAME_SIZE.
Quoted 7 lines
Yes please :) No need to respin.
Reviewed-by Philippe Mathieu-Daudé <[email protected]>Quoted 2 lines
Besides that,
Reviewed-by Marc Hartmayer <[email protected]>=
Ditto :)
Re: [PATCH v2 2/4] pc-bios/s390-ccw: Provide space for initial stack frame in start.S
In reply to Philippe Mathieu-DaudéQuoted 29 lines · Philippe Mathieu-Daudé
Quoted 18 lines · Marc Hartmayer
Quoted 14 lines · Thomas Huth
Providing the space of a stack frame is the duty of the caller, so we should reserve 160 bytes before jumping into the main function. Otherwise the
main()function might write past the stack array.While we're at it, add a proper
STACK_SIZEmacro for the stack size instead of using magic numbers (this is also required for the following patch).Reviewed-by Christian Borntraeger <[email protected]>Reviewed-by Cédric Le Goater <[email protected]>Signed-off-by Thomas Huth <[email protected]>---
pc-bios/s390-ccw/start.S| 4 +++-1 file changed, 3 insertions(+), 1 deletion(-)Quoted 9 lines
Quoted 6 lines
(quoted diff) +2
-0+#define STACK_SIZE 0x8000 +.globl
_start
_start:
- larl %r15,stack + 0x8000 /* Set up stack */ + larl %r15,stack +STACK_SIZE- 160 /* Set up stack */^^^
You can also add a macro for this
- e.g.STACK_FRAME_SIZE.
Quoted 1 line
Yes please :) No need to respin.
Ok, I'll add:
pc-bios/s390-ccw/start.S +3 -2
@@ -10,12 +10,13 @@ * directory. */ -#define STACK_SIZE 0x8000 +#define STACK_SIZE 0x8000 +#define STACK_FRAME_SIZE 160 .globl _start _start: - larl %r15,stack + STACK_SIZE - 160 /* Set up stack */ + larl %r15,stack + STACK_SIZE - STACK_FRAME_SIZE /* Set up stack */ /* clear bss */ larl %r2,.bss
Thanks,
Thomas
Re: [PATCH v2 2/4] pc-bios/s390-ccw: Provide space for initial stack frame in start.S
In reply to Thomas HuthQuoted 11 lines · On Tue, 27 Jun 2023 09:47:01 +0200 Thomas Huth <[email protected]> wrote:
Providing the space of a stack frame is the duty of the caller, so we should reserve 160 bytes before jumping into the main function. Otherwise the
main()function might write past the stack array.While we're at it, add a proper
STACK_SIZEmacro for the stack size instead of using magic numbers (this is also required for the following patch).Reviewed-by Christian Borntraeger <[email protected]>Reviewed-by Cédric Le Goater <[email protected]>Signed-off-by Thomas Huth <[email protected]>
with Marc's suggestion applied:
Quoted 22 lines
--- pc-bios/s390-ccw/start.S | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)pc-bios/s390-ccw/start.S +3
-1@@ -10,10 +10,12 @@ * directory. */ +#define STACK_SIZE 0x8000 + .globl _start _start: - larl %r15,stack + 0x8000 /* Set up stack */ + larl %r15,stack + STACK_SIZE - 160 /* Set up stack */ /* clear bss */ larl %r2,__bss_start
PATCH v2 3/4 pc-bios/s390-ccw: Move the stack array into start.S
In reply to Thomas HuthThe stack array is only referenced from the start-up code (which is shared between the s390-ccw.img and the s390-netboot.img), but it is currently declared twice, once in main.c and once in netmain.c. It makes more sense to declare this in start.S instead - which will also be helpful in the next patch, since we need to mention the .bss section in start.S in that patch.
While we're at it, let's also drop the huge alignment of the stack, since there is no technical requirement for aligning it to page boundaries.
--- pc-bios/s390-ccw/s390-ccw.h | 1 - pc-bios/s390-ccw/main.c | 1 - pc-bios/s390-ccw/netmain.c | 1 - pc-bios/s390-ccw/start.S | 6 ++++++ 4 files changed, 6 insertions(+), 3 deletions(-)
pc-bios/s390-ccw/s390-ccw.h +0 -1
@@ -55,7 +55,6 @@ void consume_io_int(void);/* main.c */ void write_subsystem_identification (void );void write_iplb_location (void ); -extern char stack[PAGE_SIZE * 8] __attribute__((__aligned__(PAGE_SIZE)));unsigned int get_loadparm_index (void );void main (void );
pc-bios/s390-ccw/main.c +0 -1
@@ -17,7 +17,6 @@#include "virtio-scsi.h" #include "dasd-ipl.h" -char stack[PAGE_SIZE * 8] __attribute__((__aligned__(PAGE_SIZE)));static SubChannelId blk_schid = { .one =1 };static char loadparm_str[LOADPARM_LEN +1 ];QemuIplParameters qipl;
pc-bios/s390-ccw/netmain.c +0 -1
@@ -50,7 +50,6 @@ void write_iplb_location(void) {}/* STSI 3.2.2 offset of first vmdb + offset of uuid inside vmdb */ #define STSI322_VMDB_UUID_OFFSET ((8 + 12) * 4) -char stack[PAGE_SIZE * 8] __attribute__((aligned(PAGE_SIZE))); IplParameterBlock iplb __attribute__((aligned(PAGE_SIZE)));static char cfgbuf[2048 ];
pc-bios/s390-ccw/start.S +6 -0
@@ -120,3 +120,9 @@ external_new_mask: .quad 0x0000000180000000 io_new_mask: .quad 0x0000000180000000 + +.bss + .align 8 +stack: + .space STACK_SIZE + .size stack,STACK_SIZE
Signature · 1 line
2.39.3
Re: [PATCH v2 3/4] pc-bios/s390-ccw: Move the stack array into start.S
Quoted 7 lines · On Tue, 27 Jun 2023 09:47:02 +0200 Thomas Huth <[email protected]> wrote:
The stack array is only referenced from the start-up code (which is shared between the s390-ccw.img and the s390-netboot.img), but it is currently declared twice, once in
main.cand once innetmain.c. It makes more sense to declare this instart.Sinstead - which will also be helpful in the next patch, since we need to mention the .bss section instart.Sin that patch.
Quoted 5 lines
While we're at it, let's also drop the huge alignment of the stack, since there is no technical requirement for aligning it to page boundaries.
Signed-off-by Thomas Huth <[email protected]>
Quoted 57 lines
--- pc-bios/s390-ccw/s390-ccw.h | 1 - pc-bios/s390-ccw/main.c | 1 - pc-bios/s390-ccw/netmain.c | 1 - pc-bios/s390-ccw/start.S | 6 ++++++ 4 files changed, 6 insertions(+), 3 deletions(-)pc-bios/s390-ccw/s390-ccw.h +0
-1@@ -55,7 +55,6 @@ void consume_io_int(void);/* main.c */ void write_subsystem_identification (void );void write_iplb_location (void ); -extern char stack[PAGE_SIZE * 8] __attribute__((__aligned__(PAGE_SIZE)));unsigned int get_loadparm_index (void );void main (void );pc-bios/s390-ccw/main.c +0
-1@@ -17,7 +17,6 @@#include "virtio-scsi.h" #include "dasd-ipl.h" -char stack[PAGE_SIZE * 8] __attribute__((__aligned__(PAGE_SIZE)));static SubChannelId blk_schid = { .one =1 };static char loadparm_str[LOADPARM_LEN +1 ];QemuIplParameters qipl;pc-bios/s390-ccw/netmain.c +0
-1@@ -50,7 +50,6 @@ void write_iplb_location(void) {}/* STSI 3.2.2 offset of first vmdb + offset of uuid inside vmdb */ #define STSI322_VMDB_UUID_OFFSET ((8 + 12) * 4) -char stack[PAGE_SIZE * 8] __attribute__((aligned(PAGE_SIZE))); IplParameterBlock iplb __attribute__((aligned(PAGE_SIZE)));static char cfgbuf[2048 ];pc-bios/s390-ccw/start.S +6
-0@@ -120,3 +120,9 @@ external_new_mask: .quad 0x0000000180000000 io_new_mask: .quad 0x0000000180000000 + +.bss + .align 8 +stack: + .space STACK_SIZE + .size stack,STACK_SIZE
Re: [PATCH v2 3/4] pc-bios/s390-ccw: Move the stack array into start.S
In reply to Thomas HuthQuoted 11 lines · Thomas Huth
The stack array is only referenced from the start-up code (which is shared between the s390-ccw.img and the s390-netboot.img), but it is currently declared twice, once in
main.cand once innetmain.c. It makes more sense to declare this instart.Sinstead - which will also be helpful in the next patch, since we need to mention the .bss section instart.Sin that patch.While we're at it, let's also drop the huge alignment of the stack, since there is no technical requirement for aligning it to page boundaries.
Quoted 7 lines
Signed-off-by Thomas Huth <[email protected]>---
pc-bios/s390-ccw/s390-ccw.h | 1 -
pc-bios/s390-ccw/main.c | 1 -
pc-bios/s390-ccw/netmain.c | 1 -
pc-bios/s390-ccw/start.S | 6 ++++++4 files changed, 6 insertions(+), 3 deletions(-)
Re: [PATCH v2 3/4] pc-bios/s390-ccw: Move the stack array into start.S
In reply to Thomas HuthQuoted 60 lines · Thomas Huth
The stack array is only referenced from the start-up code (which is shared between the s390-ccw.img and the s390-netboot.img), but it is currently declared twice, once in
main.cand once innetmain.c. It makes more sense to declare this instart.Sinstead - which will also be helpful in the next patch, since we need to mention the .bss section instart.Sin that patch.While we're at it, let's also drop the huge alignment of the stack, since there is no technical requirement for aligning it to page boundaries.
Signed-off-by Thomas Huth <[email protected]>--- pc-bios/s390-ccw/s390-ccw.h | 1 - pc-bios/s390-ccw/main.c | 1 - pc-bios/s390-ccw/netmain.c | 1 - pc-bios/s390-ccw/start.S | 6 ++++++ 4 files changed, 6 insertions(+), 3 deletions(-)pc-bios/s390-ccw/s390-ccw.h +0
-1@@ -55,7 +55,6 @@ void consume_io_int(void);/* main.c */ void write_subsystem_identification (void );void write_iplb_location (void ); -extern char stack[PAGE_SIZE * 8] __attribute__((__aligned__(PAGE_SIZE)));unsigned int get_loadparm_index (void );void main (void );pc-bios/s390-ccw/main.c +0
-1@@ -17,7 +17,6 @@#include "virtio-scsi.h" #include "dasd-ipl.h" -char stack[PAGE_SIZE * 8] __attribute__((__aligned__(PAGE_SIZE)));static SubChannelId blk_schid = { .one =1 };static char loadparm_str[LOADPARM_LEN +1 ];QemuIplParameters qipl;pc-bios/s390-ccw/netmain.c +0
-1@@ -50,7 +50,6 @@ void write_iplb_location(void) {}/* STSI 3.2.2 offset of first vmdb + offset of uuid inside vmdb */ #define STSI322_VMDB_UUID_OFFSET ((8 + 12) * 4) -char stack[PAGE_SIZE * 8] __attribute__((aligned(PAGE_SIZE))); IplParameterBlock iplb __attribute__((aligned(PAGE_SIZE)));static char cfgbuf[2048 ];pc-bios/s390-ccw/start.S +0
-0
pc-bios/s390-ccw/start.S external_new_mask: · line 128
Quoted 9 lines
pc-bios/s390-ccw/start.S +6
-0@@ -120,3 +120,9 @@ external_new_mask: .quad 0x0000000180000000 io_new_mask: .quad 0x0000000180000000 + +.bss + .align 8 +stack: + .space STACK_SIZE + .size stack,STACK_SIZE
Sigh, I missed that one of the tcg tests includes start.S ... I'll include this hunk on top of this patch to get it fixed:
tests/tcg/s390x/head64.S +2 -5
@@ -8,6 +8,8 @@ #include "../../../pc-bios/s390-ccw/start.S" #undef main +.text + main_pre: aghi %r15,-160 /* reserve stack for C code */ brasl %r14,sclp_setup
@@ -24,8 +26,3 @@ success_psw: .quad 0x2000180000000,0xfff /* see is_special_wait_psw() */ failure_psw: .quad 0x2000180000000,0 /* disabled wait */ - - .section .bss - .align 0x1000 -stack: - .skip 0x8000
Thomas
PATCH v2 4/4 pc-bios/s390-ccw: Don't use __bss_start with the "larl" instruction
In reply to Thomas Huthstart.S currently cannot be compiled with Clang 16 and binutils 2.40:
ld: start.o(.text+0x8): misaligned symbol `__bss_start' (0xc1e5) for
relocation R_390_PC32DBL
According to the built-in linker script of ld, the symbol __bss_start can actually point *before* the .bss section and does not need to have any alignment, so in certain situations (like when using the internal assembler of Clang), the __bss_start symbol can indeed be unaligned and thus it is not suitable for being used with the "larl" instruction that needs an address that is at least aligned to halfwords. The problem went unnoticed so far since binutils <= 2.39 did not check the alignment, but starting with binutils 2.40, such unaligned addresses are now refused.
Fix it by using the real start address of the .bss section instead.
--- pc-bios/s390-ccw/start.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
pc-bios/s390-ccw/start.S +1 -1
@@ -18,7 +18,7 @@ _start: larl %r15,stack + STACK_SIZE - 160 /* Set up stack */ /* clear bss */ - larl %r2,__bss_start + larl %r2,.bss larl %r3,_end slgr %r3,%r2 /* get sizeof bss */ ltgr %r3,%r3 /* bss empty? */
Signature · 1 line
2.39.3
Re: [PATCH v2 4/4] pc-bios/s390-ccw: Don't use __bss_start with the "larl" instruction
Quoted 30 lines · On Tue, 27 Jun 2023 09:47:03 +0200 Thomas Huth <[email protected]> wrote:
start.Scurrently cannot be compiled with Clang 16 and binutils 2.40:ld:
start.o(.text+0x8): misaligned symbol `__bss_start' (0xc1e5) forrelocation R_390_PC32DBLAccording to the built-in linker script of ld, the symbol
__bss_startcan actually point *before* the .bss section and does not need to have any alignment, so in certain situations (like when using the internal assembler of Clang), the__bss_startsymbol can indeed be unaligned and thus it is not suitable for being used with the "larl" instruction that needs an address that is at least aligned to halfwords. The problem went unnoticed so far since binutils <= 2.39 did not check the alignment, but starting with binutils 2.40, such unaligned addresses are now refused.Fix it by using the real start address of the .bss section instead.
Reported-by Miroslav Rezanina <[email protected]>Suggested-by Nick Clifton <[email protected]>Signed-off-by Thomas Huth <[email protected]>--- pc-bios/s390-ccw/start.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)pc-bios/s390-ccw/start.S +0
-0
pc-bios/s390-ccw/start.S _start: · line 22
Quoted 6 lines
pc-bios/s390-ccw/start.S +1
-1@@ -18,7 +18,7 @@ _start: larl %r15,stack + STACK_SIZE - 160 /* Set up stack */ /* clear bss */ - larl %r2,__bss_start + larl %r2,.bss larl %r3,_end
since we are here, do you have guarantees that _end is always correctly aligned?
if so:
Quoted 2 lines
pc-bios/s390-ccw/start.S +0
-0slgr %r3,%r2 /* get sizeof bss */ ltgr %r3,%r3 /* bss empty? */
Re: [PATCH v2 4/4] pc-bios/s390-ccw: Don't use __bss_start with the "larl" instruction
Quoted 35 lines · Claudio Imbrenda
Quoted 32 lines · On Tue, 27 Jun 2023 09:47:03 +0200 Thomas Huth <[email protected]> wrote:
start.Scurrently cannot be compiled with Clang 16 and binutils 2.40:ld:
start.o(.text+0x8): misaligned symbol `__bss_start' (0xc1e5) forrelocation R_390_PC32DBLAccording to the built-in linker script of ld, the symbol
__bss_startcan actually point *before* the .bss section and does not need to have any alignment, so in certain situations (like when using the internal assembler of Clang), the__bss_startsymbol can indeed be unaligned and thus it is not suitable for being used with the "larl" instruction that needs an address that is at least aligned to halfwords. The problem went unnoticed so far since binutils <= 2.39 did not check the alignment, but starting with binutils 2.40, such unaligned addresses are now refused.Fix it by using the real start address of the .bss section instead.
Reported-by Miroslav Rezanina <[email protected]>Suggested-by Nick Clifton <[email protected]>Signed-off-by Thomas Huth <[email protected]>--- pc-bios/s390-ccw/start.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)pc-bios/s390-ccw/start.S +0
-0@@ -18,7 +18,7 @@ _start: larl %r15,stack + STACK_SIZE - 160 /* Set up stack */
Quoted 7 lines
pc-bios/s390-ccw/start.S _start: · line 22
Quoted 4 lines
pc-bios/s390-ccw/start.S +1
-1/* clear bss */ - larl %r2,__bss_start + larl %r2,.bss larl %r3,_endsince we are here, do you have guarantees that
_endis always correctly aligned?
Yes, you can see the internal linker script by running "ld --verbose", and there I get:
...
.data1 : { *(.data1) }
_edata = .; PROVIDE (edata = .);
. = .;
__bss_start = .;
.bss :
{
*(.dynbss)
*(.bss .bss.* .gnu.linkonce.b.*)
*(COMMON)
/* Align here to ensure that the .bss section occupies space up to
_end. Align after .bss to ensure correct alignment even if the
.bss section disappears because there are no input sections.
FIXME: Why do we need it? When there is no .bss section, we do not
pad the .data section. */
. = ALIGN(. != 0 ? 64 / 8 : 1);
}
. = ALIGN(64 / 8);
. = SEGMENT_START("ldata-segment", .);
. = ALIGN(64 / 8);
_end = .; PROVIDE (end = .);
. = DATA_SEGMENT_END (.);
...As you can see, there is no alignment in front of__bss_start, but there is alignment in front of__end.
Quoted 3 lines
if so:
Reviewed-by Claudio Imbrenda <[email protected]>
Thanks!
Thomas
Re: [PATCH v2 4/4] pc-bios/s390-ccw: Don't use __bss_start with the "larl" instruction
In reply to Thomas HuthQuoted 19 lines · Thomas Huth
start.Scurrently cannot be compiled with Clang 16 and binutils 2.40:ld:
start.o(.text+0x8): misaligned symbol `__bss_start' (0xc1e5) for relocationR_390_PC32DBLAccording to the built-in linker script of ld, the symbol
__bss_startcan actually point *before* the .bss section and does not need to have
any alignment, so in certain situations (like when using the internal assembler of Clang), the__bss_startsymbol can indeed be unaligned and thus it is not suitable for being used with the "larl" instruction
that needs an address that is at least aligned to halfwords. The problem went unnoticed so far since binutils <= 2.39 did not check the alignment, but starting with binutils 2.40, such unaligned addresses are now refused.Fix it by using the real start address of the .bss section instead.
Quoted 7 lines
Reported-by Miroslav Rezanina <[email protected]>Suggested-by Nick Clifton <[email protected]>Signed-off-by Thomas Huth <[email protected]>---
pc-bios/s390-ccw/start.S | 2 +-1 file changed, 1 insertion(+), 1 deletion(-)