From: Richard Purdie <richard.pur...@linuxfoundation.org> See the patch for more details, fixes a regression in qemu causing illegal instructions in libm on powerpc, triggered by a libinput upgrade.
https://sourceware.org/git/?p=glibc.git;a=commitdiff;h=f1c56cdff09f650ad721fae026eb6a3651631f3d was the glibc code generating the instruction and triggering the issue. Signed-off-by: Richard Purdie <richard.pur...@linuxfoundation.org> Signed-off-by: Steve Sakoman <st...@sakoman.com> --- meta/recipes-devtools/qemu/qemu.inc | 1 + meta/recipes-devtools/qemu/qemu/ppc.patch | 70 +++++++++++++++++++++++ 2 files changed, 71 insertions(+) create mode 100644 meta/recipes-devtools/qemu/qemu/ppc.patch diff --git a/meta/recipes-devtools/qemu/qemu.inc b/meta/recipes-devtools/qemu/qemu.inc index e2453dd8bc..29bc34d743 100644 --- a/meta/recipes-devtools/qemu/qemu.inc +++ b/meta/recipes-devtools/qemu/qemu.inc @@ -35,6 +35,7 @@ SRC_URI = "https://download.qemu.org/${BPN}-${PV}.tar.xz \ file://0001-tracetool-use-relative-paths-for-line-preprocessor-d.patch \ file://qemu-guest-agent.init \ file://qemu-guest-agent.udev \ + file://ppc.patch \ " UPSTREAM_CHECK_REGEX = "qemu-(?P<pver>\d+(\.\d+)+)\.tar" diff --git a/meta/recipes-devtools/qemu/qemu/ppc.patch b/meta/recipes-devtools/qemu/qemu/ppc.patch new file mode 100644 index 0000000000..395cdb814f --- /dev/null +++ b/meta/recipes-devtools/qemu/qemu/ppc.patch @@ -0,0 +1,70 @@ +target/ppc: Fix fallback to MFSS for MFFSCRN, MFFSCRNI, MFFSCE and MFFSL + +The following commits changed the code such that these instructions became invalid +on pre 3.0 ISAs: + + bf8adfd88b547680aa857c46098f3a1e94373160 - target/ppc: Move mffscrn[i] to decodetree + 394c2e2fda70da722f20fb60412d6c0ca4bfaa03 - target/ppc: Move mffsce to decodetree + 3e5bce70efe6bd1f684efbb21fd2a316cbf0657e - target/ppc: Move mffsl to decodetree + +The hardware will handle them as a MFFS instruction as the code did previously. +Restore that behaviour. This means applications that were segfaulting under qemu +when encountering these instructions now operate correctly. The instruction +is used in glibc libm functions for example. + +Upstream-Status: Submitted [https://lore.kernel.org/qemu-devel/20230504110150.3044402-1-richard.pur...@linuxfoundation.org/] + +Signed-off-by: Richard Purdie <richard.pur...@linuxfoundation.org> + +Index: qemu-8.0.0/target/ppc/translate/fp-impl.c.inc +=================================================================== +--- qemu-8.0.0.orig/target/ppc/translate/fp-impl.c.inc ++++ qemu-8.0.0/target/ppc/translate/fp-impl.c.inc +@@ -584,7 +584,10 @@ static bool trans_MFFSCE(DisasContext *c + { + TCGv_i64 fpscr; + +- REQUIRE_INSNS_FLAGS2(ctx, ISA300); ++ if (unlikely(!(ctx->insns_flags2 & PPC2_ISA300))) { ++ return trans_MFFS(ctx, a); ++ } ++ + REQUIRE_FPU(ctx); + + gen_reset_fpstatus(); +@@ -597,7 +600,10 @@ static bool trans_MFFSCRN(DisasContext * + { + TCGv_i64 t1, fpscr; + +- REQUIRE_INSNS_FLAGS2(ctx, ISA300); ++ if (unlikely(!(ctx->insns_flags2 & PPC2_ISA300))) { ++ return trans_MFFS(ctx, a); ++ } ++ + REQUIRE_FPU(ctx); + + t1 = tcg_temp_new_i64(); +@@ -631,7 +637,10 @@ static bool trans_MFFSCRNI(DisasContext + { + TCGv_i64 t1, fpscr; + +- REQUIRE_INSNS_FLAGS2(ctx, ISA300); ++ if (unlikely(!(ctx->insns_flags2 & PPC2_ISA300))) { ++ return trans_MFFS(ctx, a); ++ } ++ + REQUIRE_FPU(ctx); + + t1 = tcg_temp_new_i64(); +@@ -661,7 +670,10 @@ static bool trans_MFFSCDRNI(DisasContext + { + TCGv_i64 fpscr; + +- REQUIRE_INSNS_FLAGS2(ctx, ISA300); ++ if (unlikely(!(ctx->insns_flags2 & PPC2_ISA300))) { ++ return trans_MFFS(ctx, a); ++ } ++ + REQUIRE_FPU(ctx); + + gen_reset_fpstatus(); -- 2.34.1
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#180952): https://lists.openembedded.org/g/openembedded-core/message/180952 Mute This Topic: https://lists.openembedded.org/mt/98707896/21656 Group Owner: openembedded-core+ow...@lists.openembedded.org Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-