NetCube-Systems-Austria/kumquat-buildroot
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

package/elf2flt: update RISC-V 64-bits support

Browse Source 
Remove the old elf2flt 0002-elf2flt-add-riscv-64-bits-support.patch
patch file for riscv64 architecture and replace it with 3 patches:

(1) The first patch fixes the data section alignment
(2) The second patch fixes a bug with the handling of the eh_frame
    section causing text and data section overlap problems.
(3) The third patch adds a simpler riscv64 flat bin relocation support.

These 3 patches are submitted to the upstream elf2flt project as pull
request #22:

https://github.com/uclinux-dev/elf2flt/pull/22

Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
This commit is contained in:
Damien Le Moal 2022-05-30 12:38:27 +09:00 committed by Thomas Petazzoni
parent 75301bf988
commit 32f93b0ef2
3 changed files with 181 additions and 53 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

81
package/elf2flt/0002-elf2flt.ld-reinstate-32-byte-alignment-for-.data-sec.patch Normal file
View File

@ -0,0 +1,81 @@
From 85ba5664eb368eb1cbd2c30b7cd574acd75edd4c Mon Sep 17 00:00:00 2001
From: Niklas Cassel <niklas.cassel@wdc.com>
Date: Mon, 4 Apr 2022 15:30:24 +0200
Subject: [PATCH] elf2flt.ld: reinstate 32 byte alignment for .data section
Commit 8a3e74446fe7 ("allow to build arm flat binaries") moved the
following commands:
. = ALIGN(0x20) ;
@SYMBOL_PREFIX@_etext = . ;
from the .text section to the top level in the SECTIONS node.
The .text output section is being directed to a memory region using the
"> flatmem :text" output section attribute. Commands in the top level in
the SECTIONS node are not.
This means that the ALIGN() command is no longer being appended to the
flatmem memory region, it will simply update the Location Counter.
The heuristic for placing an output section is described here:
https://sourceware.org/binutils/docs-2.38/ld.html#Output-Section-Address
"If an output memory region is set for the section then it is added to this
region and its address will be the next free address in that region."
Since the .data section is being directed to the same memory region as the
.text section, this means that the Location Counter is not used when
assigning an address to the .data output section, it will simply use the
next free address.
No longer directing these commands to the flatmem memory region means that
the .data output section is no longer aligned to a 32 byte boundary.
Before commit 8a3e74446fe7 ("allow to build arm flat binaries"):
$ readelf -S busybox_unstripped.gdb | grep data
[ 3] .data PROGBITS 0000000000035ac0 00036ac0
$ readelf -s busybox_unstripped.gdb | grep _etext
19286: 0000000000035ac0 0 NOTYPE GLOBAL DEFAULT 1 _etext
After commit 8a3e74446fe7 ("allow to build arm flat binaries"):
$ readelf -S busybox_unstripped.gdb | grep data
[ 3] .data PROGBITS 0000000000035ab0 00036ab0
$ readelf -s busybox_unstripped.gdb | grep _etext
19287: 0000000000035ac0 0 NOTYPE GLOBAL DEFAULT 3 _etext
The .data output section has to be aligned to a 32 byte boundary, see the
FLAT_DATA_ALIGN 0x20 macro and its usage in fs/binfmt_flat.c:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/fs/binfmt_flat.c?h=v5.17#n59
Readd an explicit ALIGN attribute on the .data section itself, since the
linker will obey this attribute regardless if being directed to a memory
region or not. Also remove the ALIGN() command before the .data section,
since this misleads the reader to think that the Location Counter is used
when assigning an address to the .data section, when it actually is not.
Fixes: 8a3e74446fe7 ("allow to build arm flat binaries")
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
---
elf2flt.ld.in | 5 +----
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/elf2flt.ld.in b/elf2flt.ld.in
index 0df999d..e5aea14 100644
--- a/elf2flt.ld.in
+++ b/elf2flt.ld.in
@@ -94,12 +94,9 @@ W_RODAT: *(.gnu.linkonce.r*)
*(.ARM.exidx* .gnu.linkonce.armexidx.*)
} > flatmem
@SYMBOL_PREFIX@__exidx_end = .;
-
- . = ALIGN(0x20) ;
@SYMBOL_PREFIX@_etext = . ;
- .data : {
- . = ALIGN(0x4) ;
+ .data ALIGN(0x20): {
@SYMBOL_PREFIX@_sdata = . ;
@SYMBOL_PREFIX@__data_start = . ;
@SYMBOL_PREFIX@data_start = . ;
--
2.35.1

73
package/elf2flt/0003-elf2flt-fix-.eh_frame-section-handling.patch Normal file
View File

@ -0,0 +1,73 @@
From 8b7fdb1dedfb8a6e858b46e5af33029fe0462ab8 Mon Sep 17 00:00:00 2001
From: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Date: Tue, 10 May 2022 23:14:36 +0900
Subject: [PATCH] elf2flt: fix .eh_frame section handling
elf2flt.ld linker script positions the .eh_frame section in an output
section after the .text and .data output sections.
However, when elf2flt.c is supplied the ELF linked using the elf2flt.ld
linker script, it only looks at the flags for each input section, and
then puts it in either a bFLT text, data or bss output section.
Commit ba379d08bb7 ("elf2flt: fix for segfault on some ARM ELFs")
modified the section scanning loop of elf2flt main() function to put
read-only relocation data sections in the bFLT text output section so
that the .ARM.exidx section is placed in the .text flat output section.
Previously a read-only relocation data section would be put in the data
output section.
On ARM, the .eh_frame section does not have the SEC_RELOC flag set, so
it will still end up in the data output section. However, on
architectures that generates the .eh_frame section with the SEC_RELOC
flag set, this section will now be placed in the text output section.
The logic in elf2flt will handle all sections in order, and since the
input order is .text, .data, and .eh_frame, putting .eh_frame in text
output section does not work, since elf2flt.c has already put the .data
input section in the bFLT data output section. This leads to the
following print (example for riscv64 architecture):
buildroot/output/host/riscv64-buildroot-linux-uclibc/bin/elf2flt:
ERROR: text=0x3bab8 overlaps data=0x33f60 ?
The way that elf2flt is written, we cannot append to the text output
section after an input section has been added to the data output
section. It might be possible to change this, but that would require
moving all the code the was already placed in the data output section.
Instead, let's allow putting a read-only relocation data section in the
text output section (so that .ARM.exidx will still be placed correctly),
but only if there has not yet been anything placed in the data output
section.
That way .ARM.exidx will still be placed correctly, and .eh_frame will
be placed correctly in the .data output section, regardless if it has
flag SEC_RELOC set or not.
Fixes: ba379d08bb7 ("elf2flt: fix for segfault on some ARM ELFs")
Signed-off-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
---
elf2flt.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/elf2flt.c b/elf2flt.c
index 7ac0617..da25e93 100644
--- a/elf2flt.c
+++ b/elf2flt.c
@@ -1877,8 +1877,9 @@ int main(int argc, char *argv[])
bfd_vma sec_vma;
if ((s->flags & SEC_CODE) ||
- ((s->flags & (SEC_DATA | SEC_READONLY | SEC_RELOC)) ==
- (SEC_DATA | SEC_READONLY | SEC_RELOC))) {
+ (((s->flags & (SEC_DATA | SEC_READONLY | SEC_RELOC)) ==
+ (SEC_DATA | SEC_READONLY | SEC_RELOC)) &&
+ !data_len)) {
vma = &text_vma;
len = &text_len;
} else if (s->flags & SEC_DATA) {
--
2.35.1

80
package/elf2flt/0002-elf2flt-add-riscv-64-bits-support.patch → package/elf2flt/0004-elf2flt-add-riscv-64-bits-support.patch
View File

@ -1,4 +1,4 @@
From 3879965dfda08a24e7d44ed76bbcc2f4a41df1fa Mon Sep 17 00:00:00 2001
From 3f1f323feb5cf25d8c80861991d0360784f4d2e6 Mon Sep 17 00:00:00 2001
From: Damien Le Moal <damien.lemoal@wdc.com>
Date: Wed, 9 Sep 2020 17:31:33 +0900
Subject: [PATCH] elf2flt: add riscv 64-bits support
@ -6,32 +6,28 @@ Subject: [PATCH] elf2flt: add riscv 64-bits support
Add support for riscv 64bits ISA by defining the relocation types
R_RISCV_32_PCREL, R_RISCV_ADD32, R_RISCV_SUB32, R_RISCV_32 and
R_RISCV_64. riscv64 support also needs the __global_pointer$ symbol to
be defined right after the relocation tables in the data section.
Furthermore, the .got and .got.plt sections must be reversed. These 2
requirements are handled with runtime modifications of the default
linker script using the append_sed() function.
(1) For the .got.plt and .got sections order swap, append_sed() is used
to rename "(.got.plt)" to "(.got.tmp)" and to rename "(.got)" to
"(.got.plt)". A last call finalize the name swap by replacing
"(.got.tmp)" with "(.got)"
(2) For the global pointer synbol, a definition line starting with
"RISCV_GP" is added. The "RISCV_GP" string is removed if the target CPU
type is riscv64. The definition line is dropped for other CPU types.
be defined right after the relocation tables in the data section. To
define this symbol, the "RISCV_GP" line prefix is added. The "RISCV_GP"
string is removed if the target CPU type is riscv64 and the definition
line is dropped for other CPU types.
With these changes, buildroot/busybox builds and run on NOMMU
systems with kernel 5.13. Tested on Canaan Kendryte K210 boards.
With these changes, buildroot and busybox build and run on riscv NOMMU
systems with Linux kernel including patch 6045ab5fea4c
("binfmt_flat: do not stop relocating GOT entries prematurely on riscv")
fixing the binfmt_flat loader. Tested on QEMU and Canaan Kendryte K210
boards.
This patch is based on earlier work by Christoph Hellwig <hch@lst.de>.
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
---
elf2flt.c | 23 +++++++++++++++++++++++
elf2flt.c | 16 ++++++++++++++++
elf2flt.ld.in | 1 +
ld-elf2flt.c | 16 ++++++++++++++++
3 files changed, 40 insertions(+)
ld-elf2flt.c | 8 ++++++++
3 files changed, 25 insertions(+)
diff --git a/elf2flt.c b/elf2flt.c
index ea6b5a1..7100c20 100644
index da25e93..a03ea3a 100644
--- a/elf2flt.c
+++ b/elf2flt.c
@@ -81,6 +81,8 @@ const char *elf2flt_progname;
@ -52,14 +48,16 @@ index ea6b5a1..7100c20 100644
#else
#error "Don't know how to support your CPU architecture??"
#endif
@@ -821,6 +825,16 @@ output_relocs (
@@ -812,6 +816,18 @@ output_relocs (
goto good_32bit_resolved_reloc;
default:
goto bad_resolved_reloc;
+#elif defined(TARGET_riscv64)
+ case R_RISCV_32_PCREL:
+ case R_RISCV_ADD32:
+ case R_RISCV_ADD64:
+ case R_RISCV_SUB32:
+ case R_RISCV_SUB64:
+ continue;
+ case R_RISCV_32:
+ case R_RISCV_64:
@ -69,27 +67,11 @@ index ea6b5a1..7100c20 100644
#else
default:
/* The default is to assume that the
@@ -1841,6 +1855,15 @@ int main(int argc, char *argv[])
if (!load_to_ram && !pfile)
load_to_ram = 1;
+#if defined(TARGET_riscv64)
+ /*
+ * riscv only supports loading text and data contiguously.
+ * So fail if load_to_ram is false.
+ */
+ if (!load_to_ram)
+ fatal("Loading to RAM ('-r' option) is required");
+#endif
+
fname = argv[argc-1];
if (pfile) {
diff --git a/elf2flt.ld.in b/elf2flt.ld.in
index 0df999d..f1eed1f 100644
index e5aea14..950849e 100644
--- a/elf2flt.ld.in
+++ b/elf2flt.ld.in
@@ -109,6 +109,7 @@ W_RODAT: *(.gnu.linkonce.r*)
@@ -106,6 +106,7 @@ W_RODAT: *(.gnu.linkonce.r*)
. = ALIGN(0x20) ;
LONG(-1)
. = ALIGN(0x20) ;
@ -98,32 +80,24 @@ index 0df999d..f1eed1f 100644
R_RODAT: *(.rodata1)
R_RODAT: *(.rodata.*)
diff --git a/ld-elf2flt.c b/ld-elf2flt.c
index 7cb02d5..1a503dd 100644
index 7cb02d5..75ee1bb 100644
--- a/ld-elf2flt.c
+++ b/ld-elf2flt.c
@@ -324,6 +324,22 @@ static int do_final_link(void)
@@ -324,6 +324,14 @@ static int do_final_link(void)
append_option(&other_options, concat(got_offset, "=", buf, NULL));
}
+ if (streq(TARGET_CPU, "riscv64")) {
+ /*
+ * The .got section must come before the .got.plt section
+ * (gcc/ld bug ?).
+ */
+ append_sed(&sed, "(.got.plt)", "(.got.tmp)");
+ append_sed(&sed, "(.got.plt)", "(.got)");
+ append_sed(&sed, "(.got.tmp)", "(.got.plt)");
+
+ /* The global pointer symbol is defined after the GOT. */
+ /* riscv adds a global pointer symbol to the linker file with the
+ "RISCV_GP:" prefix. Remove the prefix for riscv64 architecture and
+ the entire line for other architectures. */
+ if (streq(TARGET_CPU, "riscv64"))
+ append_sed(&sed, "^RISCV_GP:", "");
+ } else {
+ /* Get rid of the global pointer definition. */
+ else
+ append_sed(&sed, "^RISCV_GP:", NULL);
+ }
+
/* Locate the default linker script, if we don't have one provided. */
if (!linker_script)
linker_script = concat(ldscriptpath, "/elf2flt.ld", NULL);
--
2.35.1
2.36.1