Skip to contentSkip to conversation

package/gcc/9.2.0: backport or1k upstream patch

PATCH New Romain Naour 3 participants 6 messages Oct 19, 2019 – Oct 30, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 package/gcc/9.2.0: backport or1k upstream patch

Without this patch, the system build using qemu_or1k_defconfig (gcc 9.2, binutils 2.33.1 and uClibc 1.0.32) doesn't boot.

https://mailman.uclibc-ng.org/pipermail/devel/2019-August/001895.html

---
 ...-issue-with-set_got-clobbering-LR-r9.patch | 105 ++++++++++++++++++
 1 file changed, 105 insertions(+)
 create mode 100644 package/gcc/9.2.0/0003-or1k-Fix-issue-with-set_got-clobbering-LR-r9.patch
package/gcc/9.2.0/0003-or1k-Fix-issue-with-set_got-clobbering-LR-r9.patch new file +105 -0
new file mode 100644
index 0000000000..bdde2ce762
@@ -0,0 +1,105 @@
+From 94c7d630348671b781368d9270514dcf6c34e0c9 Mon Sep 17 00:00:00 2001
+From: shorne <shorne@138bc75d-0d04-0410-961f-82ee72b054a4>
+Date: Sat, 31 Aug 2019 06:00:56 +0000
+Subject: [PATCH] or1k: Fix issue with set_got clobbering LR (r9)
+
+When compiling glibc we found that the GOT register was being allocated
+r9 when the instruction was still set_got_tmp.  That is a problem
+because r9 is the Link Register (LR) in OpenRISC which is used/clobbered
+in set_got.  We cannot use r9 as the GOT register.  Also, we cannot
+simply say set_got_tmp clobbers r9 as this is the reason for having the
+temporary set_got_tmp.
+
+Fix by using a register class constraint that does not allow r9 during
+register allocation.
+
+gcc/ChangeLog:
+
+        * config/or1k/constraints.md (t): New constraint.
+        * config/or1k/or1k.h (GOT_REGS): New register class.
+        * config/or1k/or1k.md (set_got_tmp, set_got): Use t contraint.
+
+git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk at 275242 138bc75d-0d04-0410-961f-82ee72b054a4
+(cherry picked from commit 5b9471ffca25d39635680516ba2ff85295480fc3)
+Signed-off-by: Romain Naour <[email protected]>
+---
+ gcc/config/or1k/constraints.md | 4 ++++
+ gcc/config/or1k/or1k.h         | 3 +++
+ gcc/config/or1k/or1k.md        | 4 ++--
+ 3 files changed, 9 insertions(+), 2 deletions(-)
+
+diff --git a/gcc/config/or1k/constraints.md b/gcc/config/or1k/constraints.md
+index 93da8c058c6..a16b749008f 100644
+--- a/gcc/config/or1k/constraints.md
++++ b/gcc/config/or1k/constraints.md
+@@ -24,6 +24,7 @@
+ 
+ ; We use:
+ ;  c - sibcall registers
++;  t - got address registers (excludes LR (r9) which is clobbered by set_got)
+ ;  I - constant signed 16-bit
+ ;  K - constant unsigned 16-bit
+ ;  M - constant signed 16-bit shifted left 16-bits (l.movhi)
+@@ -32,6 +33,9 @@
+ (define_register_constraint "c" "SIBCALL_REGS"
+   "Registers which can hold a sibling call address")
+ 
++(define_register_constraint "t" "GOT_REGS"
++  "Registers which can be used to store the Global Offset Table (GOT) address.")
++
+ ;; Immediates
+ (define_constraint "I"
+   "A signed 16-bit immediate in the range -32768 to 32767."
+diff --git a/gcc/config/or1k/or1k.h b/gcc/config/or1k/or1k.h
+index 6dda230f217..feee702d89c 100644
+--- a/gcc/config/or1k/or1k.h
++++ b/gcc/config/or1k/or1k.h
+@@ -189,6 +189,7 @@ enum reg_class
+ {
+   NO_REGS,
+   SIBCALL_REGS,
++  GOT_REGS,
+   GENERAL_REGS,
+   FLAG_REGS,
+   ALL_REGS,
+@@ -200,6 +201,7 @@ enum reg_class
+ #define REG_CLASS_NAMES {	\
+   "NO_REGS", 			\
+   "SIBCALL_REGS",		\
++  "GOT_REGS",			\
+   "GENERAL_REGS",		\
+   "FLAG_REGS",			\
+   "ALL_REGS" }
+@@ -212,6 +214,7 @@ enum reg_class
+ #define REG_CLASS_CONTENTS      \
+ { { 0x00000000, 0x00000000 },	\
+   { SIBCALL_REGS_MASK,   0 },	\
++  { 0xfffffdff, 0x00000000 },	\
+   { 0xffffffff, 0x00000003 },	\
+   { 0x00000000, 0x00000004 },	\
+   { 0xffffffff, 0x00000007 }	\
+diff --git a/gcc/config/or1k/or1k.md b/gcc/config/or1k/or1k.md
+index 2dad51cd46b..88f3f02630f 100644
+--- a/gcc/config/or1k/or1k.md
++++ b/gcc/config/or1k/or1k.md
+@@ -595,7 +595,7 @@
+ ;; set_got pattern below.  This works because the set_got_tmp insn is the
+ ;; first insn in the stream and that it isn't moved during RA.
+ (define_insn "set_got_tmp"
+-  [(set (match_operand:SI 0 "register_operand" "=r")
++  [(set (match_operand:SI 0 "register_operand" "=t")
+ 	(unspec_volatile:SI [(const_int 0)] UNSPECV_SET_GOT))]
+   ""
+ {
+@@ -604,7 +604,7 @@
+ 
+ ;; The insn to initialize the GOT.
+ (define_insn "set_got"
+-  [(set (match_operand:SI 0 "register_operand" "=r")
++  [(set (match_operand:SI 0 "register_operand" "=t")
+ 	(unspec:SI [(const_int 0)] UNSPEC_SET_GOT))
+    (clobber (reg:SI LR_REGNUM))]
+   ""
+-- 
+2.23.0
+
Signature · 1 line
2.23.0
#1 lore raw

PATCH 2/3 package/binutils/2.32: or1k backport or1k upstream patch

Without this patch, the system build using qemu_or1k_defconfig (gcc 9.2, binutils 2.32 and uClibc 1.0.32) doesn't boot.

https://mailman.uclibc-ng.org/pipermail/devel/2019-August/001895.html

---
 ...ct-value-in-PLT-GOT-entries-causing-.patch | 46 +++++++++++++++++++
 1 file changed, 46 insertions(+)
 create mode 100644 package/binutils/2.32/0006-or1k-Fix-incorrect-value-in-PLT-GOT-entries-causing-.patch
package/binutils/2.32/0006-or1k-Fix-incorrect-value-in-PLT-GOT-entries-causing-.patch new file +46 -0
new file mode 100644
index 0000000000..8527b66733
@@ -0,0 +1,46 @@
+From b17678f639f953d687d96cd52690e7cbfae50f91 Mon Sep 17 00:00:00 2001
+From: Stafford Horne <[email protected]>
+Date: Fri, 23 Aug 2019 22:25:55 +0900
+Subject: [PATCH] or1k: Fix incorrect value in PLT GOT entries, causing
+ infinite loop
+
+The PLT GOT entry should point to the first PLT entry which contains the
+runtime linker function.  It was pointing back to the symbol PLT entry
+causing an infinite loop.
+
+I found this when testing the OpenRISC glibc port which uses the runtime
+dynamic linker.  It seems other libc's we use so far have not been
+making use of the initial PLT GOT entries.
+
+bfd/ChangeLog:
+
+	* elf32-or1k.c (or1k_elf_finish_dynamic_symbol): Use correct value for
+	PLT GOT entries.
+
+(cherry picked from commit 09f7b0de537d465fc8ed9f9433e348c1bc78aab2)
+Signed-off-by: Romain Naour <[email protected]>
+---
+ bfd/elf32-or1k.c | 7 +++++--
+ 1 file changed, 5 insertions(+), 2 deletions(-)
+
+diff --git a/bfd/elf32-or1k.c b/bfd/elf32-or1k.c
+index 2f200b197b8..32839cfa7b6 100644
+--- a/bfd/elf32-or1k.c
++++ b/bfd/elf32-or1k.c
+@@ -2379,8 +2379,11 @@ or1k_elf_finish_dynamic_symbol (bfd *output_bfd,
+       or1k_write_plt_entry (output_bfd, splt->contents + h->plt.offset,
+ 			    plt0, plt1, plt2, OR1K_JR(12));
+ 
+-      /* Fill in the entry in the global offset table.  */
+-      bfd_put_32 (output_bfd, plt_addr, sgot->contents + got_offset);
++      /* Fill in the entry in the global offset table.  We initialize it to
++	 point to the top of the plt.  This is done to lazy lookup the actual
++	 symbol as the first plt entry will be setup by libc to call the
++	 runtime dynamic linker.  */
++      bfd_put_32 (output_bfd, plt_base_addr, sgot->contents + got_offset);
+ 
+       /* Fill in the entry in the .rela.plt section.  */
+       rela.r_offset = got_addr;
+-- 
+2.23.0
+
Signature · 1 line
2.23.0
#2 lore raw

PATCH 2/3 package/binutils/2.32: or1k backport or1k upstream patch

Quoted 8 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Romain" == Romain Naour <[email protected]> writes:

Without this patch, the system build using qemu_or1k_defconfig (gcc 9.2, binutils 2.32 and uClibc 1.0.32) doesn't boot.

https://mailman.uclibc-ng.org/pipermail/devel/2019-August/001895.html

Committed to 2019.08.x, thanks.

Signature · 1 line
Bye, Peter Korsgaard
#3 lore raw

PATCH 3/3 package/gcc: or1k allow gcc 9.2 with uClibc-ng

In reply to Romain Naour

Binutils 2.32 and GCC 9.2 are now fixed thanks to Stafford Horne.

https://mailman.uclibc-ng.org/pipermail/devel/2019-August/001895.html

Fixes:
https://mailman.uclibc-ng.org/pipermail/devel/2019-August/001885.html

---
 package/gcc/Config.in.host | 3 ---
 1 file changed, 3 deletions(-)
package/gcc/Config.in.host +0 -3
index 3772125f71..1e26e747be 100644
@@ -61,9 +61,6 @@ config BR2_GCC_VERSION_9_X
 	bool "gcc 9.x"
 	# or1k needs binutils >= 2.32
 	depends on !(BR2_or1k && BR2_BINUTILS_VERSION_2_31_X)
-	# uClibc doesn't build with upstream or1k gcc
-	# https://mailman.uclibc-ng.org/pipermail/devel/2019-August/001885.html
-	depends on !(BR2_or1k && BR2_TOOLCHAIN_USES_UCLIBC)
 	# powerpc spe support has been deprecated since gcc 8.x.
 	# https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html
 	depends on !BR2_powerpc_SPE
Signature · 1 line
2.23.0
#4 lore raw

PATCH 1/3 package/gcc/9.2.0: backport or1k upstream patch

In reply to Romain Naour
Quoted 10 lines · On Sun, 20 Oct 2019 00:31:05 +0200 Romain Naour <[email protected]> wrote:

Without this patch, the system build using qemu_or1k_defconfig (gcc 9.2, binutils 2.33.1 and uClibc 1.0.32) doesn't boot.

https://mailman.uclibc-ng.org/pipermail/devel/2019-August/001895.html

---
 ...-issue-with-set_got-clobbering-LR-r9.patch | 105 ++++++++++++++++++
 1 file changed, 105 insertions(+)
 create mode 100644 package/gcc/9.2.0/0003-or1k-Fix-issue-with-set_got-clobbering-LR-r9.patch

Series applied. Note that your PATCH 3/3 doesn't apply, because it was probably based on a local patch of yours that removes binutils 2.30 (and its Config.in option), and this removal is not in master yet. I fixed the conflict when applying.

Thanks!

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
#5 lore raw

PATCH 1/3 package/gcc/9.2.0: backport or1k upstream patch

In reply to Romain Naour
Quoted 8 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Romain" == Romain Naour <[email protected]> writes:

Without this patch, the system build using qemu_or1k_defconfig (gcc 9.2, binutils 2.33.1 and uClibc 1.0.32) doesn't boot.

https://mailman.uclibc-ng.org/pipermail/devel/2019-August/001895.html

Committed to 2019.08.x, thanks. (2019.02.x does not have gcc 9.2)

Signature · 1 line
Bye, Peter Korsgaard