summaryrefslogtreecommitdiff
path: root/target/linux/brcm63xx/files/arch/mips/cfe/cfe.c
diff options
context:
space:
mode:
Diffstat (limited to 'target/linux/brcm63xx/files/arch/mips/cfe/cfe.c')
-rw-r--r--target/linux/brcm63xx/files/arch/mips/cfe/cfe.c533
1 files changed, 533 insertions, 0 deletions
diff --git a/target/linux/brcm63xx/files/arch/mips/cfe/cfe.c b/target/linux/brcm63xx/files/arch/mips/cfe/cfe.c
new file mode 100644
index 0000000000..6d16111e1c
--- /dev/null
+++ b/target/linux/brcm63xx/files/arch/mips/cfe/cfe.c
@@ -0,0 +1,533 @@
+/*
+ * Broadcom Common Firmware Environment (CFE) support
+ *
+ * Copyright 2000, 2001, 2002
+ * Broadcom Corporation. All rights reserved.
+ *
+ * Copyright (C) 2006 Michael Buesch
+ *
+ * Original Authors: Mitch Lichtenberg, Chris Demetriou
+ *
+ * This software is furnished under license and may be used and copied only
+ * in accordance with the following terms and conditions. Subject to these
+ * conditions, you may download, copy, install, use, modify and distribute
+ * modified or unmodified copies of this software in source and/or binary
+ * form. No title or ownership is transferred hereby.
+ *
+ * 1) Any source code used, modified or distributed must reproduce and
+ * retain this copyright notice and list of conditions as they appear in
+ * the source file.
+ *
+ * 2) No right is granted to use any trade name, trademark, or logo of
+ * Broadcom Corporation. The "Broadcom Corporation" name may not be
+ * used to endorse or promote products derived from this software
+ * without the prior written permission of Broadcom Corporation.
+ *
+ * 3) THIS SOFTWARE IS PROVIDED "AS-IS" AND ANY EXPRESS OR IMPLIED
+ * WARRANTIES, INCLUDING BUT NOT LIMITED TO, ANY IMPLIED WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR
+ * NON-INFRINGEMENT ARE DISCLAIMED. IN NO EVENT SHALL BROADCOM BE LIABLE
+ * FOR ANY DAMAGES WHATSOEVER, AND IN PARTICULAR, BROADCOM SHALL NOT BE
+ * LIABLE FOR DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
+ * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
+ * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+ * OR OTHERWISE), EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/init.h>
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/spinlock.h>
+#include <asm/cfe.h>
+
+#include "cfe_private.h"
+
+
+static cfe_uint_t cfe_handle;
+static int (*cfe_trampoline)(long handle, long iocb);
+
+
+#include <linux/kernel.h>
+
+void __init cfe_setup(unsigned long fwarg0, unsigned long fwarg1,
+ unsigned long fwarg2, unsigned long fwarg3)
+{
+ if (fwarg3 == 0x80300000) {
+ /* WRT54G workaround */
+ fwarg3 = CFE_EPTSEAL;
+ fwarg2 = 0xBFC00500;
+ }
+ if (fwarg3 != CFE_EPTSEAL) {
+ /* We are not booted from CFE */
+ return;
+ }
+ if (fwarg1 == 0) {
+ /* We are on the boot CPU */
+ cfe_handle = (cfe_uint_t)fwarg0;
+ cfe_trampoline = CFE_TO_PTR(fwarg2);
+ }
+}
+
+int cfe_vprintk(const char *fmt, va_list args)
+{
+ static char buffer[1024];
+ static DEFINE_SPINLOCK(lock);
+ static const char pfx[] = "CFE-console: ";
+ static const size_t pfx_len = sizeof(pfx) - 1;
+ unsigned long flags;
+ int len, cnt, pos;
+ int handle;
+ int res;
+
+ if (!cfe_present())
+ return -ENODEV;
+
+ spin_lock_irqsave(&lock, flags);
+ handle = cfe_getstdhandle(CFE_STDHANDLE_CONSOLE);
+ if (CFE_ISERR(handle)) {
+ len = -EIO;
+ goto out;
+ }
+ strcpy(buffer, pfx);
+ len = vscnprintf(buffer + pfx_len,
+ sizeof(buffer) - pfx_len - 2,
+ fmt, args);
+ len += pfx_len;
+ /* The CFE console requires CR-LF line-ends.
+ * Add a CR, if we only terminate lines with a LF.
+ * This does only fix CR-LF at the end of the string.
+ * So for multiple lines, use multiple cfe_vprintk calls.
+ */
+ if (len > 1 &&
+ buffer[len - 1] == '\n' && buffer[len - 2] != '\r') {
+ buffer[len - 1] = '\r';
+ buffer[len] = '\n';
+ len += 1;
+ }
+ cnt = len;
+ pos = 0;
+ while (cnt > 0) {
+ res = cfe_write(handle, buffer + pos, len - pos);
+ if (CFE_ISERR(res)) {
+ len = -EIO;
+ goto out;
+ }
+ cnt -= res;
+ pos += res;
+ }
+out:
+ spin_unlock_irqrestore(&lock, flags);
+
+ return len;
+}
+
+int cfe_printk(const char *fmt, ...)
+{
+ va_list args;
+ int res;
+
+ va_start(args, fmt);
+ res = cfe_vprintk(fmt, args);
+ va_end(args);
+
+ return res;
+}
+
+static int cfe_iocb_dispatch(struct cfe_iocb *iocb)
+{
+ if (!cfe_present())
+ return CFE_ERR_UNSUPPORTED;
+ return cfe_trampoline((long)cfe_handle, (long)iocb);
+}
+
+int cfe_present(void)
+{
+ return (cfe_trampoline != NULL);
+}
+
+int cfe_close(int handle)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_DEV_CLOSE;
+ iocb.handle = handle;
+
+ err = cfe_iocb_dispatch(&iocb);
+
+ return (CFE_ISERR(err)) ? err : iocb.status;
+}
+
+int cfe_cpu_start(int cpu, void (*fn)(void), long sp, long gp, long a1)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_FW_CPUCTL;
+ iocb.psize = sizeof(struct cfe_iocb_cpuctl);
+ iocb.cpuctl.number = cpu;
+ iocb.cpuctl.command = CFE_CPU_CMD_START;
+ iocb.cpuctl.gp = gp;
+ iocb.cpuctl.sp = sp;
+ iocb.cpuctl.a1 = a1;
+ iocb.cpuctl.start_addr = (long)fn;
+
+ err = cfe_iocb_dispatch(&iocb);
+
+ return (CFE_ISERR(err)) ? err : iocb.status;
+}
+
+int cfe_cpu_stop(int cpu)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_FW_CPUCTL;
+ iocb.psize = sizeof(struct cfe_iocb_cpuctl);
+ iocb.cpuctl.number = cpu;
+ iocb.cpuctl.command = CFE_CPU_CMD_STOP;
+
+ err = cfe_iocb_dispatch(&iocb);
+
+ return (CFE_ISERR(err)) ? err : iocb.status;
+}
+
+int cfe_enumenv(int idx, char *name, int namelen, char *val, int vallen)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_ENV_ENUM;
+ iocb.psize = sizeof(struct cfe_iocb_envbuf);
+ iocb.envbuf.index = idx;
+ iocb.envbuf.name = PTR_TO_CFE(name);
+ iocb.envbuf.name_len = namelen;
+ iocb.envbuf.val = PTR_TO_CFE(val);
+ iocb.envbuf.val_len = vallen;
+
+ err = cfe_iocb_dispatch(&iocb);
+
+ return (CFE_ISERR(err)) ? err : iocb.status;
+}
+
+int cfe_enumdev(int idx, char *name, int namelen)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+
+ iocb.fcode = CFE_CMD_DEV_ENUM;
+ iocb.psize = sizeof(struct cfe_iocb_envbuf);
+ iocb.envbuf.index = idx;
+ iocb.envbuf.name = PTR_TO_CFE(name);
+ iocb.envbuf.name_len = namelen;
+
+ err = cfe_iocb_dispatch(&iocb);
+
+ return (CFE_ISERR(err)) ? err : iocb.status;
+}
+
+int cfe_enummem(int idx, int flags, u64 *start, u64 *length,
+ u64 *type)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+
+ iocb.fcode = CFE_CMD_FW_MEMENUM;
+ iocb.flags = flags;
+ iocb.psize = sizeof(struct cfe_iocb_meminfo);
+ iocb.meminfo.index = idx;
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (!CFE_ISERR(iocb.status)) {
+ *start = iocb.meminfo.addr;
+ *length = iocb.meminfo.size;
+ *type = iocb.meminfo.type;
+ }
+
+ return iocb.status;
+}
+
+int cfe_exit(int warm, int status)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+printk("CFE REBOOT\n");
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_FW_RESTART;
+ if (warm)
+ iocb.flags = CFE_FLG_WARMSTART;
+ iocb.psize = sizeof(struct cfe_iocb_exitstat);
+ iocb.exitstat.status = status;
+
+printk("CALL\n");
+ err = cfe_iocb_dispatch(&iocb);
+printk("DONE\n");
+
+ return (CFE_ISERR(err)) ? err : iocb.status;
+}
+
+int cfe_flushcache(int flags)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_FW_FLUSHCACHE;
+ iocb.flags = flags;
+
+ err = cfe_iocb_dispatch(&iocb);
+
+ return (CFE_ISERR(err)) ? err : iocb.status;
+}
+
+int cfe_getdevinfo(char *name)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_DEV_GETINFO;
+ iocb.psize = sizeof(struct cfe_iocb_buf);
+ iocb.buffer.ptr = PTR_TO_CFE(name);
+ iocb.buffer.length = strlen(name);
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (CFE_ISERR(iocb.status))
+ return iocb.status;
+
+ return iocb.buffer.devflags;
+}
+
+int cfe_getenv(char *name, char *dest, int destlen)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ dest[0] = '\0';
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_ENV_GET;
+ iocb.psize = sizeof(struct cfe_iocb_envbuf);
+ iocb.envbuf.name = PTR_TO_CFE(name);
+ iocb.envbuf.name_len = strlen(name);
+ iocb.envbuf.val = PTR_TO_CFE(dest);
+ iocb.envbuf.val_len = destlen;
+
+ err = cfe_iocb_dispatch(&iocb);
+
+ return (CFE_ISERR(err)) ? err : iocb.status;
+}
+
+int cfe_getfwinfo(struct cfe_fwinfo *info)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_FW_GETINFO;
+ iocb.psize = sizeof(struct cfe_iocb_fwinfo);
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (CFE_ISERR(iocb.status))
+ return err;
+
+ info->version = iocb.fwinfo.version;
+ info->totalmem = iocb.fwinfo.totalmem;
+ info->flags = iocb.fwinfo.flags;
+ info->boardid = iocb.fwinfo.boardid;
+ info->bootarea_va = iocb.fwinfo.bootarea_va;
+ info->bootarea_pa = iocb.fwinfo.bootarea_pa;
+ info->bootarea_size = iocb.fwinfo.bootarea_size;
+
+ return iocb.status;
+}
+
+int cfe_getstdhandle(int handletype)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_DEV_GETHANDLE;
+ iocb.flags = handletype;
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (CFE_ISERR(iocb.status))
+ return iocb.status;
+
+ return iocb.handle;
+}
+
+int cfe_getticks(s64 *ticks)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_FW_GETTIME;
+ iocb.psize = sizeof(struct cfe_iocb_time);
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (!CFE_ISERR(iocb.status))
+ *ticks = iocb.time.ticks;
+
+ return iocb.status;
+}
+
+int cfe_inpstat(int handle)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_DEV_INPSTAT;
+ iocb.handle = handle;
+ iocb.psize = sizeof(struct cfe_iocb_inpstat);
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (CFE_ISERR(iocb.status))
+ return iocb.status;
+
+ return iocb.inpstat.status;
+}
+
+int cfe_ioctl(int handle, unsigned int ioctlnum,
+ unsigned char *buffer, int length,
+ int *retlen, u64 offset)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_DEV_IOCTL;
+ iocb.handle = handle;
+ iocb.psize = sizeof(struct cfe_iocb_buf);
+ iocb.buffer.offset = offset;
+ iocb.buffer.ioctlcmd = ioctlnum;
+ iocb.buffer.ptr = PTR_TO_CFE(buffer);
+ iocb.buffer.length = length;
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (CFE_ISERR(iocb.status))
+ return iocb.status;
+ if (retlen)
+ *retlen = iocb.buffer.retlen;
+
+ return iocb.status;
+}
+
+int cfe_open(char *name)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_DEV_OPEN;
+ iocb.psize = sizeof(struct cfe_iocb_buf);
+ iocb.buffer.ptr = PTR_TO_CFE(name);
+ iocb.buffer.length = strlen(name);
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (CFE_ISERR(iocb.status))
+ return iocb.status;
+
+ return iocb.handle;
+}
+
+int cfe_read(int handle, unsigned char *buffer, int length)
+{
+ return cfe_readblk(handle, 0, buffer, length);
+}
+
+int cfe_readblk(int handle, s64 offset, unsigned char *buffer, int length)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_DEV_READ;
+ iocb.handle = handle;
+ iocb.psize = sizeof(struct cfe_iocb_buf);
+ iocb.buffer.offset = offset;
+ iocb.buffer.ptr = PTR_TO_CFE(buffer);
+ iocb.buffer.length = length;
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (CFE_ISERR(iocb.status))
+ return iocb.status;
+
+ return iocb.buffer.retlen;
+}
+
+int cfe_setenv(char *name, char *val)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_ENV_SET;
+ iocb.psize = sizeof(struct cfe_iocb_envbuf);
+ iocb.envbuf.name = PTR_TO_CFE(name);
+ iocb.envbuf.name_len = strlen(name);
+ iocb.envbuf.val = PTR_TO_CFE(val);
+ iocb.envbuf.val_len = strlen(val);
+
+ err = cfe_iocb_dispatch(&iocb);
+
+ return (CFE_ISERR(err)) ? err : iocb.status;
+}
+
+int cfe_write(int handle, unsigned char *buffer, int length)
+{
+ return cfe_writeblk(handle, 0, buffer, length);
+}
+
+int cfe_writeblk(int handle, s64 offset, unsigned char *buffer, int length)
+{
+ struct cfe_iocb iocb;
+ int err;
+
+ memset(&iocb, 0, sizeof(iocb));
+ iocb.fcode = CFE_CMD_DEV_WRITE;
+ iocb.handle = handle;
+ iocb.psize = sizeof(struct cfe_iocb_buf);
+ iocb.buffer.offset = offset;
+ iocb.buffer.ptr = PTR_TO_CFE(buffer);
+ iocb.buffer.length = length;
+
+ err = cfe_iocb_dispatch(&iocb);
+ if (CFE_ISERR(err))
+ return err;
+ if (CFE_ISERR(iocb.status))
+ return iocb.status;
+
+ return iocb.buffer.retlen;
+}