Force the zero page of the kernel to be part of a segment

This will be necessary for limine and likely multiboot support. See the
comment in the linker scripts.
This commit is contained in:
bjorn3
2024-03-14 21:22:13 +01:00
parent b5f94a5a2e
commit 672e36f401
3 changed files with 38 additions and 1 deletions
+14 -1
View File
@@ -7,6 +7,19 @@ SECTIONS {
. = KERNEL_OFFSET;
. += SIZEOF_HEADERS;
/* Force the zero page to be part of a segment by creating a
* dummy section in the zero page.
* Limine will map the segment with the lowest vaddr value at
* 0xFFFFFFFF80000000 even if the segment has a higher vaddr.
* As such without the zero page being part of a segment, the
* kernel would be loaded at an offset from the expected
* location. As the redox kernel is not currently relocatable,
* this would result in a crash. A similar issue likely exists
* with multiboot/multiboot2 and the paddr of the segment.
*/
.dummy ALIGN(8) : AT(ADDR(.dummy) - KERNEL_OFFSET) {}
. = ALIGN(4096);
.text : AT(ADDR(.text) - KERNEL_OFFSET) {
@@ -21,7 +34,7 @@ SECTIONS {
__text_end = .;
}
.rodata : AT(ADDR(.rodata) - KERNEL_OFFSET) {
.rodata : AT(ADDR(.rodata) - KERNEL_OFFSET) {
__rodata_start = .;
*(.rodata*)
. = ALIGN(4096);
+12
View File
@@ -8,6 +8,18 @@ SECTIONS {
. += SIZEOF_HEADERS;
/* Force the zero page to be part of a segment by creating a
* dummy section in the zero page.
* Limine will map the segment with the lowest vaddr value at
* 0xFFFFFFFF80000000 even if the segment has a higher vaddr.
* As such without the zero page being part of a segment, the
* kernel would be loaded at an offset from the expected
* location. As the redox kernel is not currently relocatable,
* this would result in a crash. A similar issue likely exists
* with multiboot/multiboot2 and the paddr of the segment.
*/
.dummy : AT(ADDR(.dummy) - KERNEL_OFFSET) {}
.text ALIGN(4K) : AT(ADDR(.text) - KERNEL_OFFSET) {
__text_start = .;
*(.text*)
+12
View File
@@ -8,6 +8,18 @@ SECTIONS {
. += SIZEOF_HEADERS;
/* Force the zero page to be part of a segment by creating a
* dummy section in the zero page.
* Limine will map the segment with the lowest vaddr value at
* 0xFFFFFFFF80000000 even if the segment has a higher vaddr.
* As such without the zero page being part of a segment, the
* kernel would be loaded at an offset from the expected
* location. As the redox kernel is not currently relocatable,
* this would result in a crash. A similar issue likely exists
* with multiboot/multiboot2 and the paddr of the segment.
*/
.dummy : AT(ADDR(.dummy) - KERNEL_OFFSET) {}
.text ALIGN(4K) : AT(ADDR(.text) - KERNEL_OFFSET) {
__text_start = .;
*(.text*)