This patches contain the modificaion as follows:
1. Fix up the building fail on RV32.
2. Add umoddi3 and udivmoddi4 functions for RV32.
3. Fix ioremap problem on RV32.
Changes in v2:
- Retain the copyright notices from libgcc in umoddi3.c and udivmoddi4.c.
Vincent Chen (1):
RISC-V: Avoid corrupting the upper 32-bit of phys_addr_t in ioremap
Zong Li (4):
RISC-V: Build tishift only on 64-bit
RISC-V: Use swiotlb on RV64 only
lib: Add umoddi3 and udivmoddi4 of GCC library routines
RISC-V: Select GENERIC_LIB_UMODDI3 on RV32
arch/riscv/Kconfig | 1 +
arch/riscv/kernel/setup.c | 3 +
arch/riscv/lib/Makefile | 3 +-
arch/riscv/mm/ioremap.c | 2 +-
lib/Kconfig | 3 +
lib/Makefile | 1 +
lib/udivmoddi4.c | 309 ++++++++++++++++++++++++++++++++++++++++++++++
lib/umoddi3.c | 34 +++++
8 files changed, 354 insertions(+), 2 deletions(-)
create mode 100644 lib/udivmoddi4.c
create mode 100644 lib/umoddi3.c
--
2.7.4
Only RV64 supports swiotlb.
Signed-off-by: Zong Li <[email protected]>
---
arch/riscv/kernel/setup.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/arch/riscv/kernel/setup.c b/arch/riscv/kernel/setup.c
index aee6031..872a280 100644
--- a/arch/riscv/kernel/setup.c
+++ b/arch/riscv/kernel/setup.c
@@ -227,7 +227,10 @@ void __init setup_arch(char **cmdline_p)
setup_bootmem();
paging_init();
unflatten_device_tree();
+
+#ifdef CONFIG_64bit
swiotlb_init(1);
+#endif
#ifdef CONFIG_SMP
setup_smp();
--
2.7.4
On 32-bit, it need to use __umoddi3 by some drivers.
Signed-off-by: Zong Li <[email protected]>
---
arch/riscv/Kconfig | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index a344980..dc262fa 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -108,6 +108,7 @@ config ARCH_RV32I
select GENERIC_LIB_ASHRDI3
select GENERIC_LIB_LSHRDI3
select GENERIC_LIB_UCMPDI2
+ select GENERIC_LIB_UMODDI3
config ARCH_RV64I
bool "RV64I"
--
2.7.4
Add umoddi3 and udivmoddi4 support for 32-bit.
The RV32 need the umoddi3 to do modulo when the operands are long long
type, like other libraries implementation such as ucmpdi2, lshrdi3 and
so on.
I encounter the undefined reference 'umoddi3' when I use the in
house dma driver, although it is in house driver, but I think that
umoddi3 is a common function for RV32.
The udivmoddi4 and umoddi3 are copies from libgcc in gcc. There are other
functions use the udivmoddi4 in libgcc, so I separate the umoddi3 and
udivmoddi4 for flexible extension in the future.
Signed-off-by: Zong Li <[email protected]>
---
lib/Kconfig | 3 +
lib/Makefile | 1 +
lib/udivmoddi4.c | 309 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
lib/umoddi3.c | 34 ++++++
4 files changed, 347 insertions(+)
create mode 100644 lib/udivmoddi4.c
create mode 100644 lib/umoddi3.c
diff --git a/lib/Kconfig b/lib/Kconfig
index a3928d4..d82f206 100644
--- a/lib/Kconfig
+++ b/lib/Kconfig
@@ -621,3 +621,6 @@ config GENERIC_LIB_CMPDI2
config GENERIC_LIB_UCMPDI2
bool
+
+config GENERIC_LIB_UMODDI3
+ bool
diff --git a/lib/Makefile b/lib/Makefile
index ca3f7eb..51bf24d 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -271,3 +271,4 @@ obj-$(CONFIG_GENERIC_LIB_LSHRDI3) += lshrdi3.o
obj-$(CONFIG_GENERIC_LIB_MULDI3) += muldi3.o
obj-$(CONFIG_GENERIC_LIB_CMPDI2) += cmpdi2.o
obj-$(CONFIG_GENERIC_LIB_UCMPDI2) += ucmpdi2.o
+obj-$(CONFIG_GENERIC_LIB_UMODDI3) += umoddi3.o udivmoddi4.o
diff --git a/lib/udivmoddi4.c b/lib/udivmoddi4.c
new file mode 100644
index 0000000..6518b3a
--- /dev/null
+++ b/lib/udivmoddi4.c
@@ -0,0 +1,309 @@
+/*
+ * This program is free software; you can redistribute it and/or modify it under
+ * the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 3, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+ * for more details.
+ *
+ * Under Section 7 of GPL version 3, you are granted additional
+ * permissions described in the GCC Runtime Library Exception, version
+ * 3.1, as published by the Free Software Foundation.
+ *
+ * You should have received a copy of the GNU General Public License and
+ * a copy of the GCC Runtime Library Exception along with this program;
+ * see the files COPYING3 and COPYING.RUNTIME respectively.
+ */
+
+#include <linux/libgcc.h>
+
+#define count_leading_zeros(COUNT, X) ((COUNT) = __builtin_clz(X))
+
+#define W_TYPE_SIZE 32
+
+#define __ll_B ((unsigned long) 1 << (W_TYPE_SIZE / 2))
+#define __ll_lowpart(t) ((unsigned long) (t) & (__ll_B - 1))
+#define __ll_highpart(t) ((unsigned long) (t) >> (W_TYPE_SIZE / 2))
+
+/* If we still don't have umul_ppmm, define it using plain C. */
+#if !defined(umul_ppmm)
+#define umul_ppmm(w1, w0, u, v) \
+ do { \
+ unsigned long __x0, __x1, __x2, __x3; \
+ unsigned short __ul, __vl, __uh, __vh; \
+ \
+ __ul = __ll_lowpart(u); \
+ __uh = __ll_highpart(u); \
+ __vl = __ll_lowpart(v); \
+ __vh = __ll_highpart(v); \
+ \
+ __x0 = (unsigned long) __ul * __vl; \
+ __x1 = (unsigned long) __ul * __vh; \
+ __x2 = (unsigned long) __uh * __vl; \
+ __x3 = (unsigned long) __uh * __vh; \
+ \
+ __x1 += __ll_highpart(__x0); \
+ __x1 += __x2; \
+ if (__x1 < __x2) \
+ __x3 += __ll_B; \
+ \
+ (w1) = __x3 + __ll_highpart(__x1); \
+ (w0) = __ll_lowpart(__x1) * __ll_B + __ll_lowpart(__x0);\
+ } while (0)
+#endif
+
+#if !defined(sub_ddmmss)
+#define sub_ddmmss(sh, sl, ah, al, bh, bl) \
+ do { \
+ unsigned long __x; \
+ __x = (al) - (bl); \
+ (sh) = (ah) - (bh) - (__x > (al)); \
+ (sl) = __x; \
+ } while (0)
+#endif
+
+/* Define this unconditionally, so it can be used for debugging. */
+#define __udiv_qrnnd_c(q, r, n1, n0, d) \
+ do { \
+ unsigned long __d1, __d0, __q1, __q0; \
+ unsigned long __r1, __r0, __m; \
+ __d1 = __ll_highpart(d); \
+ __d0 = __ll_lowpart(d); \
+ \
+ __r1 = (n1) % __d1; \
+ __q1 = (n1) / __d1; \
+ __m = (unsigned long) __q1 * __d0; \
+ __r1 = __r1 * __ll_B | __ll_highpart(n0); \
+ if (__r1 < __m) { \
+ __q1--, __r1 += (d); \
+ if (__r1 >= (d)) \
+ if (__r1 < __m) \
+ __q1--, __r1 += (d); \
+ } \
+ __r1 -= __m; \
+ \
+ __r0 = __r1 % __d1; \
+ __q0 = __r1 / __d1; \
+ __m = (unsigned long) __q0 * __d0; \
+ __r0 = __r0 * __ll_B | __ll_lowpart(n0); \
+ if (__r0 < __m) { \
+ __q0--, __r0 += (d); \
+ if (__r0 >= (d)) \
+ if (__r0 < __m) \
+ __q0--, __r0 += (d); \
+ } \
+ __r0 -= __m; \
+ \
+ (q) = (unsigned long) __q1 * __ll_B | __q0; \
+ (r) = __r0; \
+ } while (0)
+
+/* If udiv_qrnnd was not defined for this processor, use __udiv_qrnnd_c. */
+#if !defined(udiv_qrnnd)
+#define UDIV_NEEDS_NORMALIZATION 1
+#define udiv_qrnnd __udiv_qrnnd_c
+#endif
+
+unsigned long long __udivmoddi4(unsigned long long u, unsigned long long v,
+ unsigned long long *rp)
+{
+ const DWunion nn = {.ll = u };
+ const DWunion dd = {.ll = v };
+ DWunion rr, ww;
+ unsigned long d0, d1, n0, n1, n2;
+ unsigned long q0 = 0, q1 = 0;
+ unsigned long b, bm;
+
+ d0 = dd.s.low;
+ d1 = dd.s.high;
+ n0 = nn.s.low;
+ n1 = nn.s.high;
+
+#if !UDIV_NEEDS_NORMALIZATION
+
+ if (d1 == 0) {
+ if (d0 > n1) {
+ /* 0q = nn / 0D */
+
+ udiv_qrnnd(q0, n0, n1, n0, d0);
+ q1 = 0;
+
+ /* Remainder in n0. */
+ } else {
+ /* qq = NN / 0d */
+
+ if (d0 == 0)
+ /* Divide intentionally by zero. */
+ d0 = 1 / d0;
+
+ udiv_qrnnd(q1, n1, 0, n1, d0);
+ udiv_qrnnd(q0, n0, n1, n0, d0);
+
+ /* Remainder in n0. */
+ }
+
+ if (rp != 0) {
+ rr.s.low = n0;
+ rr.s.high = 0;
+ *rp = rr.ll;
+ }
+#else /* UDIV_NEEDS_NORMALIZATION */
+
+ if (d1 == 0) {
+ if (d0 > n1) {
+ /* 0q = nn / 0D */
+
+ count_leading_zeros(bm, d0);
+
+ if (bm != 0) {
+ /*
+ * Normalize, i.e. make the most significant bit
+ * of the denominator set.
+ */
+
+ d0 = d0 << bm;
+ n1 = (n1 << bm) | (n0 >> (W_TYPE_SIZE - bm));
+ n0 = n0 << bm;
+ }
+
+ udiv_qrnnd(q0, n0, n1, n0, d0);
+ q1 = 0;
+
+ /* Remainder in n0 >> bm. */
+ } else {
+ /* qq = NN / 0d */
+
+ if (d0 == 0)
+ /* Divide intentionally by zero. */
+ d0 = 1 / d0;
+
+ count_leading_zeros(bm, d0);
+
+ if (bm == 0) {
+ /*
+ * From (n1 >= d0) /\ (the most significant bit
+ * of d0 is set), conclude (the most significant
+ * bit of n1 is set) /\ (theleading quotient
+ * digit q1 = 1).
+ *
+ * This special case is necessary, not an
+ * optimization. (Shifts counts of W_TYPE_SIZE
+ * are undefined.)
+ */
+
+ n1 -= d0;
+ q1 = 1;
+ } else {
+ /* Normalize. */
+
+ b = W_TYPE_SIZE - bm;
+
+ d0 = d0 << bm;
+ n2 = n1 >> b;
+ n1 = (n1 << bm) | (n0 >> b);
+ n0 = n0 << bm;
+
+ udiv_qrnnd(q1, n1, n2, n1, d0);
+ }
+
+ /* n1 != d0... */
+
+ udiv_qrnnd(q0, n0, n1, n0, d0);
+
+ /* Remainder in n0 >> bm. */
+ }
+
+ if (rp != 0) {
+ rr.s.low = n0 >> bm;
+ rr.s.high = 0;
+ *rp = rr.ll;
+ }
+#endif /* UDIV_NEEDS_NORMALIZATION */
+
+ } else {
+ if (d1 > n1) {
+ /* 00 = nn / DD */
+
+ q0 = 0;
+ q1 = 0;
+
+ /* Remainder in n1n0. */
+ if (rp != 0) {
+ rr.s.low = n0;
+ rr.s.high = n1;
+ *rp = rr.ll;
+ }
+ } else {
+ /* 0q = NN / dd */
+
+ count_leading_zeros(bm, d1);
+ if (bm == 0) {
+ /*
+ * From (n1 >= d1) /\ (the most significant bit
+ * of d1 is set), conclude (the most significant
+ * bit of n1 is set) /\ (the quotient digit q0 =
+ * 0 or 1).
+ *
+ * This special case is necessary, not an
+ * optimization.
+ */
+
+ /*
+ * The condition on the next line takes
+ * advantage of that n1 >= d1 (true due to
+ * program flow).
+ */
+ if (n1 > d1 || n0 >= d0) {
+ q0 = 1;
+ sub_ddmmss(n1, n0, n1, n0, d1, d0);
+ } else {
+ q0 = 0;
+ }
+
+ q1 = 0;
+
+ if (rp != 0) {
+ rr.s.low = n0;
+ rr.s.high = n1;
+ *rp = rr.ll;
+ }
+ } else {
+ unsigned long m1, m0;
+ /* Normalize. */
+
+ b = W_TYPE_SIZE - bm;
+
+ d1 = (d1 << bm) | (d0 >> b);
+ d0 = d0 << bm;
+ n2 = n1 >> b;
+ n1 = (n1 << bm) | (n0 >> b);
+ n0 = n0 << bm;
+
+ udiv_qrnnd(q0, n1, n2, n1, d1);
+ umul_ppmm(m1, m0, q0, d0);
+
+ if (m1 > n1 || (m1 == n1 && m0 > n0)) {
+ q0--;
+ sub_ddmmss(m1, m0, m1, m0, d1, d0);
+ }
+
+ q1 = 0;
+
+ /* Remainder in (n1n0 - m1m0) >> bm. */
+ if (rp != 0) {
+ sub_ddmmss(n1, n0, n1, n0, m1, m0);
+ rr.s.low = (n1 << b) | (n0 >> bm);
+ rr.s.high = n1 >> bm;
+ *rp = rr.ll;
+ }
+ }
+ }
+ }
+
+ ww.s.low = q0;
+ ww.s.high = q1;
+ return ww.ll;
+}
diff --git a/lib/umoddi3.c b/lib/umoddi3.c
new file mode 100644
index 0000000..0134696
--- /dev/null
+++ b/lib/umoddi3.c
@@ -0,0 +1,34 @@
+/*
+ * This program is free software; you can redistribute it and/or modify it under
+ * the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 3, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+ * for more details.
+ *
+ * Under Section 7 of GPL version 3, you are granted additional
+ * permissions described in the GCC Runtime Library Exception, version
+ * 3.1, as published by the Free Software Foundation.
+ *
+ * You should have received a copy of the GNU General Public License and
+ * a copy of the GCC Runtime Library Exception along with this program;
+ * see the files COPYING3 and COPYING.RUNTIME respectively.
+ */
+
+#include <linux/module.h>
+#include <linux/libgcc.h>
+
+extern unsigned long long __udivmoddi4(unsigned long long u,
+ unsigned long long v,
+ unsigned long long *rp);
+
+unsigned long long __umoddi3(unsigned long long u, unsigned long long v)
+{
+ unsigned long long w;
+ (void)__udivmoddi4(u, v, &w);
+ return w;
+}
+EXPORT_SYMBOL(__umoddi3);
--
2.7.4
Only RV64 supports 128 integer size.
Signed-off-by: Zong Li <[email protected]>
---
arch/riscv/lib/Makefile | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/arch/riscv/lib/Makefile b/arch/riscv/lib/Makefile
index 445ec84..5739bd0 100644
--- a/arch/riscv/lib/Makefile
+++ b/arch/riscv/lib/Makefile
@@ -2,6 +2,7 @@ lib-y += delay.o
lib-y += memcpy.o
lib-y += memset.o
lib-y += uaccess.o
-lib-y += tishift.o
+
+lib-(CONFIG_64BIT) += tishift.o
lib-$(CONFIG_32BIT) += udivdi3.o
--
2.7.4
From: Vincent Chen <[email protected]>
For 32bit, the upper 32-bit of phys_addr_t will be flushed to zero
after AND with PAGE_MASK because the data type of PAGE_MASK is
unsigned long. To fix this problem, the page alignment is done by
subtracting the page offset instead of AND with PAGE_MASK.
Signed-off-by: Vincent Chen <[email protected]>
---
arch/riscv/mm/ioremap.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/riscv/mm/ioremap.c b/arch/riscv/mm/ioremap.c
index 70ef272..bd2f2db 100644
--- a/arch/riscv/mm/ioremap.c
+++ b/arch/riscv/mm/ioremap.c
@@ -42,7 +42,7 @@ static void __iomem *__ioremap_caller(phys_addr_t addr, size_t size,
/* Page-align mappings */
offset = addr & (~PAGE_MASK);
- addr &= PAGE_MASK;
+ addr -= offset;
size = PAGE_ALIGN(size + offset);
area = get_vm_area_caller(size, VM_IOREMAP, caller);
--
2.7.4
On Wed, Sep 26, 2018 at 04:31:12PM +0800, Zong Li wrote:
> diff --git a/lib/udivmoddi4.c b/lib/udivmoddi4.c
> new file mode 100644
> index 0000000..6518b3a
> --- /dev/null
> +++ b/lib/udivmoddi4.c
> @@ -0,0 +1,309 @@
needs and SPDX tag.
> +/*
> + * This program is free software; you can redistribute it and/or modify it under
> + * the terms of the GNU General Public License as published by the Free
> + * Software Foundation; either version 3, or (at your option) any later
> + * version.
We can't take GPLv3 code for the kernel.
> + * Under Section 7 of GPL version 3, you are granted additional
> + * permissions described in the GCC Runtime Library Exception, version
> + * 3.1, as published by the Free Software Foundation.
> + *
> + * You should have received a copy of the GNU General Public License and
> + * a copy of the GCC Runtime Library Exception along with this program;
> + * see the files COPYING3 and COPYING.RUNTIME respectively.
unless this exception somehow counts (which I doubt), but for that we'd
actually need it in the tree.
On Wed, 26 Sep 2018 01:31:09 PDT (-0700), [email protected] wrote:
> This patches contain the modificaion as follows:
> 1. Fix up the building fail on RV32.
> 2. Add umoddi3 and udivmoddi4 functions for RV32.
> 3. Fix ioremap problem on RV32.
>
> Changes in v2:
> - Retain the copyright notices from libgcc in umoddi3.c and udivmoddi4.c.
>
> Vincent Chen (1):
> RISC-V: Avoid corrupting the upper 32-bit of phys_addr_t in ioremap
>
> Zong Li (4):
> RISC-V: Build tishift only on 64-bit
> RISC-V: Use swiotlb on RV64 only
> lib: Add umoddi3 and udivmoddi4 of GCC library routines
> RISC-V: Select GENERIC_LIB_UMODDI3 on RV32
>
> arch/riscv/Kconfig | 1 +
> arch/riscv/kernel/setup.c | 3 +
> arch/riscv/lib/Makefile | 3 +-
> arch/riscv/mm/ioremap.c | 2 +-
> lib/Kconfig | 3 +
> lib/Makefile | 1 +
> lib/udivmoddi4.c | 309 ++++++++++++++++++++++++++++++++++++++++++++++
> lib/umoddi3.c | 34 +++++
> 8 files changed, 354 insertions(+), 2 deletions(-)
> create mode 100644 lib/udivmoddi4.c
> create mode 100644 lib/umoddi3.c
I like what this does, but there's a license issue here that Christoph pointed
out. It'd be great to have RV32I systems booting 4.19, do you think there's a
chance you can sort this out next week so we can get it into the last RC? That
would help with the rv32i glibc process, as we'll actually be able to boot an
upstream kernel.
It looks like the functions we have an issue with come from libgcc2.c, which
doesn't change very much. As far as I can tell we might be able to get away
with these functions from 4.2.1, which was licensed under the GPLv2. Here's
the relevant functions on GCC's github mirror
https://github.com/gcc-mirror/gcc/blob/gcc-4_2_1-release/gcc/libgcc2.c#L559
https://github.com/gcc-mirror/gcc/blob/gcc-4_2_1-release/gcc/libgcc2.c#L844
These look very similar to what you send out. As far as I can tell, if you
base your code on the GPLv2 version then this should be fine to include.
Thanks!
Palmer Dabbelt <[email protected]> 於 2018年9月29日 週六 上午7:52寫道：
>
> On Wed, 26 Sep 2018 01:31:09 PDT (-0700), [email protected] wrote:
> > This patches contain the modificaion as follows:
> > 1. Fix up the building fail on RV32.
> > 2. Add umoddi3 and udivmoddi4 functions for RV32.
> > 3. Fix ioremap problem on RV32.
> >
> > Changes in v2:
> > - Retain the copyright notices from libgcc in umoddi3.c and udivmoddi4.c.
> >
> > Vincent Chen (1):
> > RISC-V: Avoid corrupting the upper 32-bit of phys_addr_t in ioremap
> >
> > Zong Li (4):
> > RISC-V: Build tishift only on 64-bit
> > RISC-V: Use swiotlb on RV64 only
> > lib: Add umoddi3 and udivmoddi4 of GCC library routines
> > RISC-V: Select GENERIC_LIB_UMODDI3 on RV32
> >
> > arch/riscv/Kconfig | 1 +
> > arch/riscv/kernel/setup.c | 3 +
> > arch/riscv/lib/Makefile | 3 +-
> > arch/riscv/mm/ioremap.c | 2 +-
> > lib/Kconfig | 3 +
> > lib/Makefile | 1 +
> > lib/udivmoddi4.c | 309 ++++++++++++++++++++++++++++++++++++++++++++++
> > lib/umoddi3.c | 34 +++++
> > 8 files changed, 354 insertions(+), 2 deletions(-)
> > create mode 100644 lib/udivmoddi4.c
> > create mode 100644 lib/umoddi3.c
>
> I like what this does, but there's a license issue here that Christoph pointed
> out. It'd be great to have RV32I systems booting 4.19, do you think there's a
> chance you can sort this out next week so we can get it into the last RC? That
> would help with the rv32i glibc process, as we'll actually be able to boot an
> upstream kernel.
>
> It looks like the functions we have an issue with come from libgcc2.c, which
> doesn't change very much. As far as I can tell we might be able to get away
> with these functions from 4.2.1, which was licensed under the GPLv2. Here's
> the relevant functions on GCC's github mirror
>
> https://github.com/gcc-mirror/gcc/blob/gcc-4_2_1-release/gcc/libgcc2.c#L559
> https://github.com/gcc-mirror/gcc/blob/gcc-4_2_1-release/gcc/libgcc2.c#L844
>
> These look very similar to what you send out. As far as I can tell, if you
> base your code on the GPLv2 version then this should be fine to include.
>
> Thanks!
Thanks for the help and useful information. I will modify the
implementation and submit the next version.