char *arc_cache_mumbojumbo(int c, char *buf, int len)
{
int n = 0;
struct cpuinfo_arc_cache *p;
#define IS_USED_RUN(v) ((v) ? "" : "(disabled) ")
#define PR_CACHE(p, cfg, str) \
if (!(p)->ver) \
n += scnprintf(buf + n, len - n, str"\t\t: N/A\n"); \
else \
n += scnprintf(buf + n, len - n, \
str"\t\t: %uK, %dway/set, %uB Line, %s%s%s\n", \
(p)->sz_k, (p)->assoc, (p)->line_len, \
(p)->vipt ? "VIPT" : "PIPT", \
(p)->alias ? " aliasing" : "", \
IS_ENABLED(cfg) ? "" : " (not used)");
PR_CACHE(&cpuinfo_arc700[c].icache, CONFIG_ARC_HAS_ICACHE, "I-Cache");
PR_CACHE(&cpuinfo_arc700[c].dcache, CONFIG_ARC_HAS_DCACHE, "D-Cache");
if (!is_isa_arcv2())
return buf;
p = &cpuinfo_arc700[c].slc;
if (p->ver)
n += scnprintf(buf + n, len - n,
"SLC\t\t: %uK, %uB Line%s\n",
p->sz_k, p->line_len, IS_USED_RUN(slc_enable));
if (ioc_exists)
n += scnprintf(buf + n, len - n, "IOC\t\t:%s\n",
IS_USED_RUN(ioc_enable));
return buf;
}
/*
* Common routine to print scratch regs (r0-r12) or callee regs (r13-r25)
* -Prints 3 regs per line and a CR.
* -To continue, callee regs right after scratch, special handling of CR
*/
static noinline void print_reg_file(long *reg_rev, int start_num)
{
unsigned int i;
char buf[512];
int n = 0, len = sizeof(buf);
for (i = start_num; i < start_num + 13; i++) {
n += scnprintf(buf + n, len - n, "r%02u: 0x%08lx\t",
i, (unsigned long)*reg_rev);
if (((i + 1) % 3) == 0)
n += scnprintf(buf + n, len - n, "\n");
/* because pt_regs has regs reversed: r12..r0, r25..r13 */
if (is_isa_arcv2() && start_num == 0)
reg_rev++;
else
reg_rev--;
}
if (start_num != 0)
n += scnprintf(buf + n, len - n, "\n\n");
/* To continue printing callee regs on same line as scratch regs */
if (start_num == 0)
pr_info("%s", buf);
else
pr_cont("%s\n", buf);
}
void read_decode_cache_bcr(void)
{
struct cpuinfo_arc_cache *p_ic, *p_dc;
unsigned int cpu = smp_processor_id();
struct bcr_cache {
#ifdef CONFIG_CPU_BIG_ENDIAN
unsigned int pad:12, line_len:4, sz:4, config:4, ver:8;
#else
unsigned int ver:8, config:4, sz:4, line_len:4, pad:12;
#endif
} ibcr, dbcr;
p_ic = &cpuinfo_arc700[cpu].icache;
READ_BCR(ARC_REG_IC_BCR, ibcr);
if (!ibcr.ver)
goto dc_chk;
if (ibcr.ver <= 3) {
BUG_ON(ibcr.config != 3);
p_ic->assoc = 2; /* Fixed to 2w set assoc */
} else if (ibcr.ver >= 4) {
p_ic->assoc = 1 << ibcr.config; /* 1,2,4,8 */
}
p_ic->line_len = 8 << ibcr.line_len;
p_ic->sz_k = 1 << (ibcr.sz - 1);
p_ic->ver = ibcr.ver;
p_ic->vipt = 1;
p_ic->alias = p_ic->sz_k/p_ic->assoc/TO_KB(PAGE_SIZE) > 1;
dc_chk:
p_dc = &cpuinfo_arc700[cpu].dcache;
READ_BCR(ARC_REG_DC_BCR, dbcr);
if (!dbcr.ver)
goto slc_chk;
if (dbcr.ver <= 3) {
BUG_ON(dbcr.config != 2);
p_dc->assoc = 4; /* Fixed to 4w set assoc */
p_dc->vipt = 1;
p_dc->alias = p_dc->sz_k/p_dc->assoc/TO_KB(PAGE_SIZE) > 1;
} else if (dbcr.ver >= 4) {
p_dc->assoc = 1 << dbcr.config; /* 1,2,4,8 */
p_dc->vipt = 0;
p_dc->alias = 0; /* PIPT so can't VIPT alias */
}
p_dc->line_len = 16 << dbcr.line_len;
p_dc->sz_k = 1 << (dbcr.sz - 1);
p_dc->ver = dbcr.ver;
slc_chk:
if (is_isa_arcv2())
read_decode_cache_bcr_arcv2(cpu);
}
static char *arc_extn_mumbojumbo(int cpu_id, char *buf, int len)
{
int n = 0;
struct cpuinfo_arc *cpu = &cpuinfo_arc700[cpu_id];
FIX_PTR(cpu);
n += scnprintf(buf + n, len - n, "Vector Table\t: %#x\n", cpu->vec_base);
if (cpu->extn.fpu_sp || cpu->extn.fpu_dp)
n += scnprintf(buf + n, len - n, "FPU\t\t: %s%s\n",
IS_AVAIL1(cpu->extn.fpu_sp, "SP "),
IS_AVAIL1(cpu->extn.fpu_dp, "DP "));
if (cpu->extn.ap_num | cpu->extn.smart | cpu->extn.rtt) {
n += scnprintf(buf + n, len - n, "DEBUG\t\t: %s%s",
IS_AVAIL1(cpu->extn.smart, "smaRT "),
IS_AVAIL1(cpu->extn.rtt, "RTT "));
if (cpu->extn.ap_num) {
n += scnprintf(buf + n, len - n, "ActionPoint %d/%s",
cpu->extn.ap_num,
cpu->extn.ap_full ? "full":"min");
}
n += scnprintf(buf + n, len - n, "\n");
}
if (cpu->dccm.sz || cpu->iccm.sz)
n += scnprintf(buf + n, len - n, "Extn [CCM]\t: DCCM @ %x, %d KB / ICCM: @ %x, %d KB\n",
cpu->dccm.base_addr, TO_KB(cpu->dccm.sz),
cpu->iccm.base_addr, TO_KB(cpu->iccm.sz));
if (is_isa_arcv2()) {
/* Error Protection: ECC/Parity */
struct bcr_erp erp;
READ_BCR(ARC_REG_ERP_BUILD, erp);
if (erp.ver) {
struct ctl_erp ctl;
READ_BCR(ARC_REG_ERP_CTRL, ctl);
/* inverted bits: 0 means enabled */
n += scnprintf(buf + n, len - n, "Extn [ECC]\t: %s%s%s%s%s%s\n",
IS_AVAIL3(erp.ic, !ctl.dpi, "IC "),
IS_AVAIL3(erp.dc, !ctl.dpd, "DC "),
IS_AVAIL3(erp.mmu, !ctl.mpd, "MMU "));
}
}
n += scnprintf(buf + n, len - n, "OS ABI [v%d]\t: %s\n",
EF_ARC_OSABI_CURRENT >> 8,
EF_ARC_OSABI_CURRENT == EF_ARC_OSABI_V3 ?
"no-legacy-syscalls" : "64-bit data any register aligned");
return buf;
}
void dma_free_coherent(struct device *dev, size_t size, void *kvaddr,
dma_addr_t dma_handle)
{
if (is_isa_arcv2() && ioc_exists)
return dma_free_noncoherent(dev, size, kvaddr, dma_handle);
iounmap((void __force __iomem *)kvaddr);
free_pages_exact((void *)dma_handle, size);
}
void arc_cache_init(void)
{
unsigned int __maybe_unused cpu = smp_processor_id();
char str[256];
printk(arc_cache_mumbojumbo(0, str, sizeof(str)));
if (IS_ENABLED(CONFIG_ARC_HAS_ICACHE)) {
struct cpuinfo_arc_cache *ic = &cpuinfo_arc700[cpu].icache;
if (!ic->ver)
panic("cache support enabled but non-existent cache\n");
if (ic->line_len != L1_CACHE_BYTES)
panic("ICache line [%d] != kernel Config [%d]",
ic->line_len, L1_CACHE_BYTES);
if (ic->ver != CONFIG_ARC_MMU_VER)
panic("Cache ver [%d] doesn't match MMU ver [%d]\n",
ic->ver, CONFIG_ARC_MMU_VER);
/*
* In MMU v4 (HS38x) the alising icache config uses IVIL/PTAG
* pair to provide vaddr/paddr respectively, just as in MMU v3
*/
if (is_isa_arcv2() && ic->alias)
_cache_line_loop_ic_fn = __cache_line_loop_v3;
else
_cache_line_loop_ic_fn = __cache_line_loop;
}
if (IS_ENABLED(CONFIG_ARC_HAS_DCACHE)) {
struct cpuinfo_arc_cache *dc = &cpuinfo_arc700[cpu].dcache;
if (!dc->ver)
panic("cache support enabled but non-existent cache\n");
if (dc->line_len != L1_CACHE_BYTES)
panic("DCache line [%d] != kernel Config [%d]",
dc->line_len, L1_CACHE_BYTES);
/* check for D-Cache aliasing on ARCompact: ARCv2 has PIPT */
if (is_isa_arcompact()) {
int handled = IS_ENABLED(CONFIG_ARC_CACHE_VIPT_ALIASING);
if (dc->alias && !handled)
panic("Enable CONFIG_ARC_CACHE_VIPT_ALIASING\n");
else if (!dc->alias && handled)
panic("Disable CONFIG_ARC_CACHE_VIPT_ALIASING\n");
}
}
}
static void arc_chk_core_config(void)
{
struct cpuinfo_arc *cpu = &cpuinfo_arc700[smp_processor_id()];
int fpu_enabled;
if (!cpu->extn.timer0)
panic("Timer0 is not present!\n");
if (!cpu->extn.timer1)
panic("Timer1 is not present!\n");
if (IS_ENABLED(CONFIG_ARC_HAS_RTC) && !cpu->extn.rtc)
panic("RTC is not present\n");
#ifdef CONFIG_ARC_HAS_DCCM
/*
* DCCM can be arbit placed in hardware.
* Make sure it's placement/sz matches what Linux is built with
*/
if ((unsigned int)__arc_dccm_base != cpu->dccm.base_addr)
panic("Linux built with incorrect DCCM Base address\n");
if (CONFIG_ARC_DCCM_SZ != cpu->dccm.sz)
panic("Linux built with incorrect DCCM Size\n");
#endif
#ifdef CONFIG_ARC_HAS_ICCM
if (CONFIG_ARC_ICCM_SZ != cpu->iccm.sz)
panic("Linux built with incorrect ICCM Size\n");
#endif
/*
* FP hardware/software config sanity
* -If hardware contains DPFP, kernel needs to save/restore FPU state
* -If not, it will crash trying to save/restore the non-existant regs
*
* (only DPDP checked since SP has no arch visible regs)
*/
fpu_enabled = IS_ENABLED(CONFIG_ARC_FPU_SAVE_RESTORE);
if (cpu->extn.fpu_dp && !fpu_enabled)
pr_warn("CONFIG_ARC_FPU_SAVE_RESTORE needed for working apps\n");
else if (!cpu->extn.fpu_dp && fpu_enabled)
panic("FPU non-existent, disable CONFIG_ARC_FPU_SAVE_RESTORE\n");
if (is_isa_arcv2() && IS_ENABLED(CONFIG_SMP) && cpu->isa.atomic &&
IS_ENABLED(CONFIG_ARC_HAS_LLSC) &&
!IS_ENABLED(CONFIG_ARC_STAR_9000923308))
panic("llock/scond livelock workaround missing\n");
}
static void arc_dma_free(struct device *dev, size_t size, void *vaddr,
dma_addr_t dma_handle, struct dma_attrs *attrs)
{
struct page *page = virt_to_page(dma_handle);
int is_non_coh = 1;
is_non_coh = dma_get_attr(DMA_ATTR_NON_CONSISTENT, attrs) ||
(is_isa_arcv2() && ioc_exists);
if (PageHighMem(page) || !is_non_coh)
iounmap((void __force __iomem *)vaddr);
__free_pages(page, get_order(size));
}
void arch_dma_free(struct device *dev, size_t size, void *vaddr,
dma_addr_t dma_handle, unsigned long attrs)
{
phys_addr_t paddr = dma_handle;
struct page *page = virt_to_page(paddr);
int is_non_coh = 1;
is_non_coh = (attrs & DMA_ATTR_NON_CONSISTENT) ||
(is_isa_arcv2() && ioc_enable);
if (PageHighMem(page) || !is_non_coh)
iounmap((void __force __iomem *)vaddr);
__free_pages(page, get_order(size));
}
void *dma_alloc_coherent(struct device *dev, size_t size,
dma_addr_t *dma_handle, gfp_t gfp)
{
void *paddr, *kvaddr;
/*
* IOC relies on all data (even coherent DMA data) being in cache
* Thus allocate normal cached memory
*
* The gains with IOC are two pronged:
* -For streaming data, elides needs for cache maintenance, saving
* cycles in flush code, and bus bandwidth as all the lines of a
* buffer need to be flushed out to memory
* -For coherent data, Read/Write to buffers terminate early in cache
* (vs. always going to memory - thus are faster)
*/
if (is_isa_arcv2() && ioc_exists)
return dma_alloc_noncoherent(dev, size, dma_handle, gfp);
/* This is linear addr (0x8000_0000 based) */
paddr = alloc_pages_exact(size, gfp);
if (!paddr)
return NULL;
/* This is kernel Virtual address (0x7000_0000 based) */
kvaddr = ioremap_nocache((unsigned long)paddr, size);
if (kvaddr == NULL)
return NULL;
/* This is bus address, platform dependent */
*dma_handle = (dma_addr_t)paddr;
/*
* Evict any existing L1 and/or L2 lines for the backing page
* in case it was used earlier as a normal "cached" page.
* Yeah this bit us - STAR 9000898266
*
* Although core does call flush_cache_vmap(), it gets kvaddr hence
* can't be used to efficiently flush L1 and/or L2 which need paddr
* Currently flush_cache_vmap nukes the L1 cache completely which
* will be optimized as a separate commit
*/
dma_cache_wback_inv((unsigned long)paddr, size);
return kvaddr;
}
void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
gfp_t gfp, unsigned long attrs)
{
unsigned long order = get_order(size);
struct page *page;
phys_addr_t paddr;
void *kvaddr;
int need_coh = 1, need_kvaddr = 0;
page = alloc_pages(gfp, order);
if (!page)
return NULL;
/*
* IOC relies on all data (even coherent DMA data) being in cache
* Thus allocate normal cached memory
*
* The gains with IOC are two pronged:
* -For streaming data, elides need for cache maintenance, saving
* cycles in flush code, and bus bandwidth as all the lines of a
* buffer need to be flushed out to memory
* -For coherent data, Read/Write to buffers terminate early in cache
* (vs. always going to memory - thus are faster)
*/
if ((is_isa_arcv2() && ioc_enable) ||
(attrs & DMA_ATTR_NON_CONSISTENT))
need_coh = 0;
/*
* - A coherent buffer needs MMU mapping to enforce non-cachability
* - A highmem page needs a virtual handle (hence MMU mapping)
* independent of cachability
*/
if (PageHighMem(page) || need_coh)
need_kvaddr = 1;
/* This is linear addr (0x8000_0000 based) */
paddr = page_to_phys(page);
*dma_handle = paddr;
/* This is kernel Virtual address (0x7000_0000 based) */
if (need_kvaddr) {
kvaddr = ioremap_nocache(paddr, size);
if (kvaddr == NULL) {
__free_pages(page, order);
return NULL;
}
} else {
kvaddr = (void *)(u32)paddr;
}
/*
* Evict any existing L1 and/or L2 lines for the backing page
* in case it was used earlier as a normal "cached" page.
* Yeah this bit us - STAR 9000898266
*
* Although core does call flush_cache_vmap(), it gets kvaddr hence
* can't be used to efficiently flush L1 and/or L2 which need paddr
* Currently flush_cache_vmap nukes the L1 cache completely which
* will be optimized as a separate commit
*/
if (need_coh)
dma_cache_wback_inv(paddr, size);
return kvaddr;
}
static inline int need_slc_flush(void)
{
return is_isa_arcv2() && l2_line_sz;
}
static char *arc_cpu_mumbojumbo(int cpu_id, char *buf, int len)
{
struct cpuinfo_arc *cpu = &cpuinfo_arc700[cpu_id];
struct bcr_identity *core = &cpu->core;
int i, n = 0, ua = 0;
FIX_PTR(cpu);
n += scnprintf(buf + n, len - n,
"\nIDENTITY\t: ARCVER [%#02x] ARCNUM [%#02x] CHIPID [%#4x]\n",
core->family, core->cpu_id, core->chip_id);
n += scnprintf(buf + n, len - n, "processor [%d]\t: %s %s (%s ISA) %s%s%s\n",
cpu_id, cpu->name, cpu->details,
is_isa_arcompact() ? "ARCompact" : "ARCv2",
IS_AVAIL1(cpu->isa.be, "[Big-Endian]"),
IS_AVAIL3(cpu->extn.dual, cpu->extn.dual_enb, " Dual-Issue "));
n += scnprintf(buf + n, len - n, "Timers\t\t: %s%s%s%s%s%s\nISA Extn\t: ",
IS_AVAIL1(cpu->extn.timer0, "Timer0 "),
IS_AVAIL1(cpu->extn.timer1, "Timer1 "),
IS_AVAIL2(cpu->extn.rtc, "RTC [UP 64-bit] ", CONFIG_ARC_TIMERS_64BIT),
IS_AVAIL2(cpu->extn.gfrc, "GFRC [SMP 64-bit] ", CONFIG_ARC_TIMERS_64BIT));
#ifdef __ARC_UNALIGNED__
ua = 1;
#endif
n += i = scnprintf(buf + n, len - n, "%s%s%s%s%s%s",
IS_AVAIL2(cpu->isa.atomic, "atomic ", CONFIG_ARC_HAS_LLSC),
IS_AVAIL2(cpu->isa.ldd, "ll64 ", CONFIG_ARC_HAS_LL64),
IS_AVAIL1(cpu->isa.unalign, "unalign "), IS_USED_RUN(ua));
if (i)
n += scnprintf(buf + n, len - n, "\n\t\t: ");
if (cpu->extn_mpy.ver) {
if (cpu->extn_mpy.ver <= 0x2) { /* ARCompact */
n += scnprintf(buf + n, len - n, "mpy ");
} else {
int opt = 2; /* stock MPY/MPYH */
if (cpu->extn_mpy.dsp) /* OPT 7-9 */
opt = cpu->extn_mpy.dsp + 6;
n += scnprintf(buf + n, len - n, "mpy[opt %d] ", opt);
}
}
n += scnprintf(buf + n, len - n, "%s%s%s%s%s%s%s%s\n",
IS_AVAIL1(cpu->isa.div_rem, "div_rem "),
IS_AVAIL1(cpu->extn.norm, "norm "),
IS_AVAIL1(cpu->extn.barrel, "barrel-shift "),
IS_AVAIL1(cpu->extn.swap, "swap "),
IS_AVAIL1(cpu->extn.minmax, "minmax "),
IS_AVAIL1(cpu->extn.crc, "crc "),
IS_AVAIL2(cpu->extn.swape, "swape", CONFIG_ARC_HAS_SWAPE));
if (cpu->bpu.ver)
n += scnprintf(buf + n, len - n,
"BPU\t\t: %s%s match, cache:%d, Predict Table:%d",
IS_AVAIL1(cpu->bpu.full, "full"),
IS_AVAIL1(!cpu->bpu.full, "partial"),
cpu->bpu.num_cache, cpu->bpu.num_pred);
if (is_isa_arcv2()) {
struct bcr_lpb lpb;
READ_BCR(ARC_REG_LPB_BUILD, lpb);
if (lpb.ver) {
unsigned int ctl;
ctl = read_aux_reg(ARC_REG_LPB_CTRL);
n += scnprintf(buf + n, len - n, " Loop Buffer:%d %s",
lpb.entries,
IS_DISABLED_RUN(!ctl));
}
}
n += scnprintf(buf + n, len - n, "\n");
return buf;
}