From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from bombadil.infradead.org (bombadil.infradead.org [198.137.202.133]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id C83ACC433EF for ; Mon, 14 Feb 2022 20:03:08 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=lists.infradead.org; s=bombadil.20210309; h=Sender: Content-Transfer-Encoding:Content-Type:List-Subscribe:List-Help:List-Post: List-Archive:List-Unsubscribe:List-Id:MIME-Version:References:In-Reply-To: Message-ID:Date:Subject:Cc:To:From:Reply-To:Content-ID:Content-Description: Resent-Date:Resent-From:Resent-Sender:Resent-To:Resent-Cc:Resent-Message-ID: List-Owner; bh=uzvXE2b3guHvPx0bjO76HZJE79J7l+vDNPa60F4ic1E=; b=B1rr6WpN8uv677 VL72au8QkT6MqT7Vc7962ykFtqnZZdyV4Il4rxhnSwauAh5it4DDuoT+M7mTaKUkgWKFuv00s/0nR akRpv6uQmNFYqz5OWWURm5JwJLGqHb6XFp7bccJJc5egZXvxZ2eOzqrYaWiY0EjzegH2BNVBAeEVG RE8Et/nLp0RK8a37k0bv+5RDn4VSlh4Y0g9YNVo2qR1u6Mof6mBYqu81qwEPkU6KgaJFy+R0yNvyo fpFDXiJ76BehETKBmfJbnUk35206WTNdValqH3B/0Zt9enO378nLe0Xh1vSjmQ9Y1ATfmSO30VVXK 7emUW8LiPDKA3qG8JBjg==; Received: from localhost ([::1] helo=bombadil.infradead.org) by bombadil.infradead.org with esmtp (Exim 4.94.2 #2 (Red Hat Linux)) id 1nJhYo-00Gm2O-1k; Mon, 14 Feb 2022 20:02:58 +0000 Received: from gloria.sntech.de ([185.11.138.130]) by bombadil.infradead.org with esmtps (Exim 4.94.2 #2 (Red Hat Linux)) id 1nJhYk-00Gm0i-4L for linux-riscv@lists.infradead.org; Mon, 14 Feb 2022 20:02:56 +0000 Received: from ip5b412258.dynamic.kabel-deutschland.de ([91.65.34.88] helo=diego.localnet) by gloria.sntech.de with esmtpsa (TLS1.3:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.92) (envelope-from ) id 1nJhYY-0002ir-VS; Mon, 14 Feb 2022 21:02:43 +0100 From: Heiko =?ISO-8859-1?Q?St=FCbner?= To: Atish Patra Cc: Palmer Dabbelt , Paul Walmsley , Albert Ou , linux-riscv , devicetree , "linux-kernel@vger.kernel.org List" , Rob Herring , Wei Fu , liush , Guo Ren , Anup Patel , Drew Fustini , Christoph Hellwig , Arnd Bergmann , Chen-Yu Tsai , Maxime Ripard , Greg Favor , Andrea Mondelli , Jonathan Behrens , Xinhaoqu , Bill Huffman , Nick Kossifidis , Allen Baum , Josh Scheid , Richard Trauben , Samuel Holland , Christoph Muellner , Philipp Tomsich Subject: Re: [PATCH v6 00/14] riscv: support for Svpbmt and D1 memory types Date: Mon, 14 Feb 2022 21:02:41 +0100 Message-ID: <1644870918.0lLKBYk3Mk@diego> In-Reply-To: References: <20220209123800.269774-1-heiko@sntech.de> <2177281.3HUnQTRebA@diego> MIME-Version: 1.0 X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.8.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20220214_120254_497773_D5D6F880 X-CRM114-Status: GOOD ( 51.48 ) X-BeenThere: linux-riscv@lists.infradead.org X-Mailman-Version: 2.1.34 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Sender: "linux-riscv" Errors-To: linux-riscv-bounces+linux-riscv=archiver.kernel.org@lists.infradead.org Hi Atish, Am Samstag, 12. Februar 2022, 01:25:53 CET schrieb Atish Patra: > On Thu, Feb 10, 2022 at 6:04 PM Heiko St=FCbner wrote: > > > > Am Freitag, 11. Februar 2022, 02:48:38 CET schrieb Atish Patra: > > > On Thu, Feb 10, 2022 at 4:25 PM Atish Patra w= rote: > > > > > > > > On Wed, Feb 9, 2022 at 4:38 AM Heiko Stuebner wro= te: > > > > > > > > > > Svpbmt is an extension defining "Supervisor-mode: page-based memo= ry types" > > > > > for things like non-cacheable pages or I/O memory pages. > > > > > > > > > > > > > > > So this is my 2nd try at implementing Svpbmt (and the diverging D= 1 memory > > > > > types) using the alternatives framework. > > > > > > > > > > This includes a number of changes to the alternatives mechanism i= tself. > > > > > The biggest one being the move to a more central location, as I e= xpect > > > > > in the future, nearly every chip needing some sort of patching, b= e it > > > > > either for erratas or for optional features (svpbmt or others). > > > > > > > > > > The dt-binding for svpbmt itself is of course not finished and is= still > > > > > using the binding introduced in previous versions, as where to put > > > > > a svpbmt-property in the devicetree is still under dicussion. > > > > > Atish seems to be working on a framework for extensions [0], > > > > > > > > > > > > > Here is the patch series > > > > https://lore.kernel.org/lkml/20220210214018.55739-1-atishp@rivosinc= .com/ > > > > > > > > I think we can simplify the cpu feature probing in PATCH 10 with the > > > > above series > > > > which simply relies on the existing riscv_isa bitmap. > > > > > > > > We also don't need the separate svpbmt property in DT mmu node. > > > > Let me know what you think. > > > > > > > > > The series also introduces support for the memory types of the D1 > > > > > which are implemented differently to svpbmt. But when patching an= yway > > > > > it's pretty clean to add the D1 variant via ALTERNATIVE_2 to the = same > > > > > location. > > > > > > > > > > The only slightly bigger difference is that the "normal" type is = not 0 > > > > > as with svpbmt, so kernel patches for this PMA type need to be ap= plied > > > > > even before the MMU is brought up, so the series introduces a sep= arate > > > > > stage for that. > > > > > > > > > > > > > > > In theory this series is 3 parts: > > > > > - sbi cache-flush / null-ptr > > > > > - alternatives improvements > > > > > - svpbmt+d1 > > > > > > > > > > So expecially patches from the first 2 areas could be applied when > > > > > deemed ready, I just thought to keep it together to show-case whe= re > > > > > the end-goal is and not requiring jumping between different serie= s. > > > > > > > > > > > > > > > The sbi cache-flush patch is based on Atish's sparse-hartid patch= [1], > > > > > as it touches a similar area in mm/cacheflush.c > > > > > > > > > > > > > > > I picked the recipient list from the previous version, hopefully > > > > > I didn't forget anybody. > > > > > > > > > > > I am also getting a load access fault while booting this series in Qe= mu. > > > > > > > > > sbi_trap_error_debug: hart1: trap handler failed (error -2) > > > sbi_trap_error_debug: hart1: mcause=3D0x0000000000000005 mtval=3D0x00= 00000080046468 > > > sbi_trap_error_debug: hart1: mtval2=3D0x0000000000000000 mtinst=3D0x0= 000000000000000 > > > sbi_trap_error_debug: hart1: mepc=3D0x000000008080a8b8 mstatus=3D0x00= 00000a00000800 > > > sbi_trap_error_debug: hart1: ra=3D0x0000000080202b06 sp=3D0x000000008= 1203f00 > > > sbi_trap_error_debug: hart1: gp=3D0x00000000812d9db8 tp=3D0x000000008= 0046000 > > > sbi_trap_error_debug: hart1: s0=3D0x0000000081203f80 s1=3D0x000000008= 0c1a8a8 > > > sbi_trap_error_debug: hart1: a0=3D0x0000000080c1a8a8 a1=3D0x000000008= 0c1b0d0 > > > sbi_trap_error_debug: hart1: a2=3D0x0000000000000002 a3=3D0x000000000= 0000000 > > > sbi_trap_error_debug: hart1: a4=3D0x00000000812da902 a5=3D0x000000000= 0000000 > > > sbi_trap_error_debug: hart1: a6=3D0x0000000000000006 a7=3D0x000000000= 0000010 > > > sbi_trap_error_debug: hart1: s2=3D0x0000000080c1b0d0 s3=3D0x000000000= 0000002 > > > sbi_trap_error_debug: hart1: s4=3D0x00000000bf000000 s5=3D0x000000000= 0000000 > > > sbi_trap_error_debug: hart1: s6=3D0x8000000a00006800 s7=3D0x000000000= 000007f > > > sbi_trap_error_debug: hart1: s8=3D0x0000000080018038 s9=3D0x000000008= 0039eac > > > sbi_trap_error_debug: hart1: s10=3D0x0000000000000000 s11=3D0x0000000= 000000000 > > > sbi_trap_error_debug: hart1: t0=3D0x0000000080c04000 t1=3D0x000000000= 0000002 > > > sbi_trap_error_debug: hart1: t2=3D0x0000000000001000 t3=3D0x000000000= 0000010 > > > sbi_trap_error_debug: hart1: t4=3D0x00000000800168be t5=3D0x000000000= 0000027 > > > sbi_trap_error_debug: hart1: t6=3D0x0000000000000001 > > > > > > mepc : 0x000000008080a8b8 - call_function_init (kernel/smp.c) > > > > > > Kernel - 5.17-rc2 + my patches > > > Qemu - Alistairs next tree + my patches > > > > very strange. I was testing of course with Qemu as well, though never s= aw > > anything like this. > > > > But of course it was Qemu master + the then still pending svpbmt patchs= et [0] > > [looks like Alistair applied this today] + a patch that made qemu inser= t the > > svpbmt dt-property for the virt machine. > > > > Oh ... just to make sure, did you enable the svpbmt parameter when star= ting > > Qemu? (-cpu ...,svpbmt=3Dtrue) > > > = > Yeah. I tried with or without. It's failing in both cases. I found a > fix but that may be unrelated and hiding the real issue. > Marking the cpufeature_svpbmt_check_of functions inline allows me to boot. > = > Here is my analysis: > = > Here is the trace of the trap: > sbi_trap_error_debug: hart0: trap handler failed (error -2) > sbi_trap_error_debug: hart0: mcause=3D0x0000000000000005 mtval=3D0x000000= 0080048468 > sbi_trap_error_debug: hart0: mtval2=3D0x0000000000000000 mtinst=3D0x00000= 00000000000 > sbi_trap_error_debug: hart0: mepc=3D0x000000008080a8b8 mstatus=3D0x000000= 0a00000800 > sbi_trap_error_debug: hart0: ra=3D0x0000000080202b06 sp=3D0x0000000081203= f00 > sbi_trap_error_debug: hart0: gp=3D0x00000000812d9db8 tp=3D0x0000000080048= 000 > sbi_trap_error_debug: hart0: s0=3D0x0000000081203f80 s1=3D0x0000000080c1a= 8a8 > sbi_trap_error_debug: hart0: a0=3D0x0000000080c1a8a8 a1=3D0x0000000080c1b= 0d0 > sbi_trap_error_debug: hart0: a2=3D0x0000000000000002 a3=3D0x0000000000000= 000 > sbi_trap_error_debug: hart0: a4=3D0x00000000812da902 a5=3D0x0000000000000= 000 > sbi_trap_error_debug: hart0: a6=3D0x0000000000000006 a7=3D0x0000000000000= 010 > sbi_trap_error_debug: hart0: s2=3D0x0000000080c1b0d0 s3=3D0x0000000000000= 002 > sbi_trap_error_debug: hart0: s4=3D0x00000000bf000000 s5=3D0x0000000000000= 000 > sbi_trap_error_debug: hart0: s6=3D0x8000000a00006800 s7=3D0x0000000000000= 07f > sbi_trap_error_debug: hart0: s8=3D0x0000000080018038 s9=3D0x0000000080039= ea8 > sbi_trap_error_debug: hart0: s10=3D0x0000000000000000 s11=3D0x00000000000= 00000 > sbi_trap_error_debug: hart0: t0=3D0x0000000080c04000 t1=3D0x0000000000000= 002 > sbi_trap_error_debug: hart0: t2=3D0x0000000000001000 t3=3D0x0000000000000= 010 > sbi_trap_error_debug: hart0: t4=3D0x00000000800168be t5=3D0x0000000000000= 027 > sbi_trap_error_debug: hart0: t6=3D0x0000000000000001 > = > mepc : 0x000000008080a8b8 - should be ffffffff8060a8b8 in objdump > output after offset > = > Here is the snippet of the objdump output for riscv_cpufeature_patch_func > = > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > inline output: (booting fails with above dump) > = > void riscv_cpufeature_patch_func(struct alt_entry *begin, struct alt_entr= y *end, > unsigned int stage) > { > ffffffff8060a89a: 7119 addi sp,sp,-128 > ffffffff8060a89c: f8a2 sd s0,112(sp) > ffffffff8060a89e: f4a6 sd s1,104(sp) > ffffffff8060a8a0: f0ca sd s2,96(sp) > ffffffff8060a8a2: e4d6 sd s5,72(sp) > ffffffff8060a8a4: fc86 sd ra,120(sp) > ffffffff8060a8a6: ecce sd s3,88(sp) > ffffffff8060a8a8: e8d2 sd s4,80(sp) > ffffffff8060a8aa: e0da sd s6,64(sp) > ffffffff8060a8ac: fc5e sd s7,56(sp) > ffffffff8060a8ae: f862 sd s8,48(sp) > ffffffff8060a8b0: f466 sd s9,40(sp) > ffffffff8060a8b2: f06a sd s10,32(sp) > ffffffff8060a8b4: ec6e sd s11,24(sp) > ffffffff8060a8b6: 0100 addi s0,sp,128 > ffffffff8060a8b8: 46823783 ld a5,1128(tp) # > 468 <__efistub_.L0 +0x5> --------> Faulting instruction > ffffffff8060a8bc: f8f43423 sd a5,-120(s0) > ffffffff8060a8c0: 4781 li a5,0 > ffffffff8060a8c2: 8932 mv s2,a2 > ffffffff8060a8c4: 84aa mv s1,a0 > ffffffff8060a8c6: 8aae mv s5,a1 > switch (stage) { > ffffffff8060a8c8: ca1d beqz > a2,ffffffff8060a8fe > ffffffff8060a8ca: 4789 li a5,2 > ffffffff8060a8cc: 18f60363 beq > a2,a5,ffffffff8060aa52 > for_each_of_cpu_node(node) { > ffffffff8060a8d0: 4501 li a0,0 > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > = > noinline output (boots fine) > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > void riscv_cpufeature_patch_func(struct alt_entry *begin, struct alt_entr= y *end, > unsigned int stage) > { > ffffffff8060a968: 7159 addi sp,sp,-112 > ffffffff8060a96a: f0a2 sd s0,96(sp) > ffffffff8060a96c: eca6 sd s1,88(sp) > ffffffff8060a96e: e8ca sd s2,80(sp) > ffffffff8060a970: fc56 sd s5,56(sp) > ffffffff8060a972: f486 sd ra,104(sp) > ffffffff8060a974: e4ce sd s3,72(sp) > ffffffff8060a976: e0d2 sd s4,64(sp) > ffffffff8060a978: f85a sd s6,48(sp) > ffffffff8060a97a: f45e sd s7,40(sp) > ffffffff8060a97c: f062 sd s8,32(sp) > ffffffff8060a97e: ec66 sd s9,24(sp) > ffffffff8060a980: e86a sd s10,16(sp) > ffffffff8060a982: e46e sd s11,8(sp) > ffffffff8060a984: 1880 addi s0,sp,112 > ffffffff8060a986: 8932 mv s2,a2 > ffffffff8060a988: 84aa mv s1,a0 > ffffffff8060a98a: 8aae mv s5,a1 > switch (stage) { > ffffffff8060a98c: ca09 beqz > a2,ffffffff8060a99e > ffffffff8060a98e: 4789 li a5,2 > ffffffff8060a990: 0ef60f63 beq > a2,a5,ffffffff8060aa8e > return cpufeature_svpbmt_check_of(); > ffffffff8060a994: f07ff0ef jal > ra,ffffffff8060a89a > cpu_req_feature |=3D (1U << idx); > ffffffff8060a998: 0005091b sext.w s2,a0 > ffffffff8060a99c: a8d5 j > ffffffff8060aa90 > const void *fdt =3D dtb_early_va; > = > = > Any thoughts ? It looks like it may related to "tp" Faulting instruction "efistub", so maybe something between efi and the devicetree not agreeing? Though at least on my build I also have efistub enabled, so it should be the same. So it's very strange that you're seeing that trap, which I've never seen so far. In any case, I tried your + Tsukasa's patches regarding the isa-extensions today, and obviously that works fine, so we wil get rid of the devicetree- parts anyway I guess. With the change below on top of that v2/v3 version-mix, I also get a working svpbmt of course. After hacking qemu to generate a suitable isa-string for me :-) . Heiko -------------- 8< ------------------ diff --git a/arch/riscv/include/asm/hwcap.h b/arch/riscv/include/asm/hwcap.h index 691fc9c8099b..656cd626eb1a 100644 --- a/arch/riscv/include/asm/hwcap.h +++ b/arch/riscv/include/asm/hwcap.h @@ -51,6 +51,7 @@ extern unsigned long elf_hwcap; * available logical extension id. */ enum riscv_isa_ext_id { + RISCV_ISA_EXT_SVPBMT =3D RISCV_ISA_EXT_BASE, RISCV_ISA_EXT_ID_MAX =3D RISCV_ISA_EXT_MAX, }; = diff --git a/arch/riscv/kernel/cpu.c b/arch/riscv/kernel/cpu.c index ced7e5be8641..b5130b15ca8d 100644 --- a/arch/riscv/kernel/cpu.c +++ b/arch/riscv/kernel/cpu.c @@ -71,6 +71,7 @@ int riscv_of_parent_hartid(struct device_node *node) } = static struct riscv_isa_ext_data isa_ext_arr[] =3D { + __RISCV_ISA_EXT_DATA("svpbmt", RISCV_ISA_EXT_SVPBMT), __RISCV_ISA_EXT_DATA("", RISCV_ISA_EXT_MAX), }; = diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c index 2e4eaedbf7f5..d82033ece1fd 100644 --- a/arch/riscv/kernel/cpufeature.c +++ b/arch/riscv/kernel/cpufeature.c @@ -192,6 +192,8 @@ void __init riscv_fill_hwcap(void) if (!ext_long) { this_hwcap |=3D isa2hwcap[(unsigned char)(*ext)]; this_isa |=3D (1UL << (*ext - 'a')); + } else { + SET_ISA_EXT_MAP("svpbmt", RISCV_ISA_EXT_SVPBMT); } #undef SET_ISA_EXT_MAP } @@ -253,64 +255,6 @@ struct cpufeature_info { bool (*check_func)(unsigned int stage); }; = -#if defined(CONFIG_MMU) && defined(CONFIG_64BIT) -static bool __init_or_module cpufeature_svpbmt_check_fdt(void) -{ - const void *fdt =3D dtb_early_va; - const char *str; - int offset; - - offset =3D fdt_path_offset(fdt, "/cpus"); - if (offset < 0) - return false; - - for (offset =3D fdt_next_node(fdt, offset, NULL); offset >=3D 0; - offset =3D fdt_next_node(fdt, offset, NULL)) { - str =3D fdt_getprop(fdt, offset, "device_type", NULL); - if (!str || strcmp(str, "cpu")) - break; - - str =3D fdt_getprop(fdt, offset, "mmu-type", NULL); - if (!str) - continue; - - if (!strncmp(str + 6, "none", 4)) - continue; - - str =3D fdt_getprop(fdt, offset, "riscv,mmu", NULL); - if (!str) - continue; - - if (!strncmp(str + 6, "svpbmt", 6)) - return true; - } - - return false; -} - -static bool __init_or_module cpufeature_svpbmt_check_of(void) -{ - struct device_node *node; - const char *str; - - for_each_of_cpu_node(node) { - if (of_property_read_string(node, "mmu-type", &str)) - continue; - - if (!strncmp(str + 6, "none", 4)) - continue; - - if (of_property_read_string(node, "riscv,mmu", &str)) - continue; - - if (!strncmp(str + 6, "svpbmt", 6)) - return true; - } - - return false; -} -#endif - static bool __init_or_module cpufeature_svpbmt_check_func(unsigned int sta= ge) { bool ret =3D false; @@ -319,10 +263,8 @@ static bool __init_or_module cpufeature_svpbmt_check_f= unc(unsigned int stage) switch (stage) { case RISCV_ALTERNATIVES_EARLY_BOOT: return false; - case RISCV_ALTERNATIVES_BOOT: - return cpufeature_svpbmt_check_fdt(); default: - return cpufeature_svpbmt_check_of(); + return riscv_isa_extension_available(NULL, SVPBMT); } #endif = diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S index c01012f3740d..ec07f991866a 100644 --- a/arch/riscv/kernel/head.S +++ b/arch/riscv/kernel/head.S @@ -10,7 +10,6 @@ #include #include #include -#include #include #include #include @@ -341,7 +340,6 @@ clear_bss_done: call kasan_early_init #endif /* Start the kernel */ - call apply_boot_alternatives call soc_early_init tail start_kernel = diff --git a/arch/riscv/kernel/setup.c b/arch/riscv/kernel/setup.c index 339ceb595b38..b4879c942b42 100644 --- a/arch/riscv/kernel/setup.c +++ b/arch/riscv/kernel/setup.c @@ -21,6 +21,7 @@ #include #include = +#include #include #include #include @@ -295,6 +296,7 @@ void __init setup_arch(char **cmdline_p) #endif = riscv_fill_hwcap(); + apply_boot_alternatives(); } = static int __init topology_init(void) _______________________________________________ linux-riscv mailing list linux-riscv@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-riscv