/* * Mostly platform independent upcall operations to Venus: * -- upcalls * -- upcall routines * * Linux 2.0 version * Copyright (C) 1996 Peter J. Braam , * Michael Callahan * * Redone for Linux 2.1 * Copyright (C) 1997 Carnegie Mellon University * * Carnegie Mellon University encourages users of this code to contribute * improvements to the Coda project. Contact Peter Braam . */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static vcsize = (sizeof(struct inputArgs) > sizeof(struct outputArgs)) ? sizeof(struct inputArgs): sizeof(struct outputArgs); /* the upcalls */ int venus_rootfid(struct super_block *sb, ViceFid *fidp) { struct inputArgs *inp; struct outputArgs *outp; int error=0; int size; ENTRY; UPARG(vcsize, CFS_ROOT); error = coda_upcall(coda_sbp(sb), VC_IN_NO_DATA, &size, inp); if (error) { printk("coda_get_rootfid: error %d\n", error); } else { *fidp = (ViceFid) outp->d.cfs_root.VFid; CDEBUG(D_SUPER, "VolumeId: %lx, VnodeId: %lx.\n", fidp->Volume, fidp->Vnode); } if (inp) CODA_FREE(inp, VC_IN_NO_DATA); EXIT; return -error; } int venus_getattr(struct super_block *sb, struct ViceFid *fid, struct coda_vattr *attr) { struct inputArgs *inp; struct outputArgs *outp; int size, error; ENTRY; UPARG(vcsize, CFS_GETATTR); inp->d.cfs_getattr.VFid = *fid; error = coda_upcall(coda_sbp(sb), vcsize, &size, inp); if ( !error ) *attr = (struct coda_vattr) outp->d.cfs_getattr.attr; if (inp) CODA_FREE(inp, sizeof(struct inputArgs)); EXIT; return -error; } int venus_setattr(struct super_block *sb, struct ViceFid *fid, struct coda_vattr *vattr) { struct inputArgs *inp; struct outputArgs *outp; int error, size; UPARG(vcsize, CFS_SETATTR); inp->d.cfs_setattr.VFid = *fid; inp->d.cfs_setattr.attr = *vattr; error = coda_upcall(coda_sbp(sb), vcsize, &size, inp); CDEBUG(D_SUPER, " result %ld\n", outp->result); if ( inp ) CODA_FREE(inp, vcsize); return -error; } int venus_lookup(struct super_block *sb, struct ViceFid *fid, const char *name, int length, int * type, struct ViceFid *resfid) { struct inputArgs *inp; struct outputArgs *outp; int insize, size, error=0, payload_offset; insize = VC_INSIZE(cfs_lookup_in) + CFS_MAXNAMLEN +1; UPARG(insize, CFS_LOOKUP); inp->d.cfs_lookup.VFid = *fid; /* send Venus a null terminated string */ payload_offset = VC_INSIZE(cfs_lookup_in); inp->d.cfs_lookup.name = (char *) payload_offset; memcpy((char *)inp + payload_offset, name, length); *((char *)inp + payload_offset + length) = '\0'; size = payload_offset + length + 1; error = coda_upcall(coda_sbp(sb), size, &size, inp); if ( !error ) { *resfid = outp->d.cfs_lookup.VFid; *type = outp->d.cfs_lookup.vtype; } if (inp) CODA_FREE(inp, insize); return -error; } int venus_release(struct super_block *sb, struct ViceFid *fid, int flags) { struct inputArgs *inp; struct outputArgs *outp; int size = sizeof(struct outputArgs); int error = 0; CODA_ALLOC(inp, struct inputArgs *, sizeof(struct inputArgs)); outp = (struct outputArgs *)inp; INIT_IN(inp, CFS_CLOSE); coda_load_creds(&(inp->cred)); inp->d.cfs_close.VFid = *fid; inp->d.cfs_close.flags = flags; error = coda_upcall(coda_sbp(sb), size, &size, inp); if (inp) CODA_FREE(inp, sizeof(struct inputArgs)); return -error; } int venus_open(struct super_block *sb, struct ViceFid *fid, int flags, ino_t *ino, dev_t *dev) { struct inputArgs *inp = NULL; struct outputArgs *outp = NULL; int size = sizeof(struct inputArgs); int error = 0; CODA_ALLOC(inp, struct inputArgs *, sizeof(struct inputArgs)); outp = (struct outputArgs *)inp; INIT_IN(inp, CFS_OPEN); coda_load_creds(&(inp->cred)); inp->d.cfs_open.VFid = *fid; inp->d.cfs_open.flags = flags; error = coda_upcall(coda_sbp(sb), size, &size, inp); if ( !error ) { *ino = outp->d.cfs_open.inode; *dev = outp->d.cfs_open.dev; } else { *ino = 0; *dev = 0; } if (inp) CODA_FREE(inp, sizeof(struct inputArgs)); return -error; } int venus_mkdir(struct super_block *sb, struct ViceFid *dirfid, const char *name, int length, struct ViceFid *newfid, struct coda_vattr *attrs) { struct inputArgs *inp; struct outputArgs *outp; int error=0, size, payload_offset; payload_offset = VC_INSIZE(cfs_mkdir_in); size = CFS_MAXNAMLEN + payload_offset; UPARG(size, CFS_MKDIR); inp->d.cfs_mkdir.VFid = *dirfid; inp->d.cfs_mkdir.attr = *attrs; inp->d.cfs_mkdir.name = (char *) payload_offset; /* Venus must get null terminated string */ memcpy((char *)inp + payload_offset, name, length); *((char *)inp + payload_offset + length) = '\0'; size = payload_offset + length + 1; error = coda_upcall(coda_sbp(sb), size, &size, inp); *attrs = outp->d.cfs_mkdir.attr; *newfid = outp->d.cfs_mkdir.VFid; if (inp) CODA_FREE(inp, size); return -error; } int venus_rename(struct super_block *sb, struct ViceFid *old_fid, struct ViceFid *new_fid, size_t old_length, size_t new_length, const char *old_name, const char *new_name) { struct inputArgs *inp; struct outputArgs *outp; int error, offset, size, s; size = 2*CFS_MAXNAMLEN + VC_INSIZE(cfs_rename_in) +8; UPARG(size, CFS_RENAME); inp->d.cfs_rename.sourceFid = *old_fid; inp->d.cfs_rename.destFid = *new_fid; offset = VC_INSIZE(cfs_rename_in); /* Venus must receive an null terminated string */ inp->d.cfs_rename.srcname = (char *)offset; s = ( old_length & ~0x3) +4; /* round up to word boundary */ memcpy((char *)inp + offset, old_name, old_length); *((char *)inp + offset + old_length) = '\0'; /* another null terminated string for Venus */ offset += s; inp->d.cfs_rename.destname = (char *)offset; s = ( new_length & ~0x3) +4; /* round up to word boundary */ memcpy((char *)inp + offset, new_name, new_length); *((char *)inp + offset + new_length) = '\0'; size += s; CDEBUG(D_INODE, "destname in packet: %s\n", (char *)inp + (int) inp->d.cfs_rename.destname); error = coda_upcall(coda_sbp(sb), size, &size, inp); if (inp) CODA_FREE(inp, size); return -error; } int venus_create(struct super_block *sb, struct ViceFid *dirfid, const char *name, int length, int excl, int mode, struct ViceFid *newfid, struct coda_vattr *attrs) { struct inputArgs *inp; struct outputArgs *outp; int error=0, size, payload_offset; payload_offset = VC_INSIZE(cfs_create_in); size = CFS_MAXNAMLEN + payload_offset; UPARG(size, CFS_CREATE); inp->d.cfs_create.VFid = *dirfid; inp->d.cfs_create.attr.va_mode = mode; inp->d.cfs_create.excl = excl; inp->d.cfs_create.mode = mode; inp->d.cfs_create.name = (char *) payload_offset; /* Venus must get null terminated string */ memcpy((char *)inp + payload_offset, name, length); *((char *)inp + payload_offset + length) = '\0'; size = payload_offset + length + 1; error = coda_upcall(coda_sbp(sb), size, &size, inp); *attrs = outp->d.cfs_create.attr; *newfid = outp->d.cfs_create.VFid; if (inp) CODA_FREE(inp, size); return -error; } int venus_rmdir(struct super_block *sb, struct ViceFid *dirfid, const char *name, int length) { struct inputArgs *inp; struct outputArgs *outp; int error=0, size, payload_offset; payload_offset = VC_INSIZE(cfs_rmdir_in); size = CFS_MAXNAMLEN + payload_offset; UPARG(size, CFS_RMDIR); inp->d.cfs_rmdir.VFid = *dirfid; inp->d.cfs_rmdir.name = (char *) payload_offset; memcpy((char *)inp + payload_offset, name, size); error = coda_upcall(coda_sbp(sb), size, &size, inp); if ( inp ) CODA_FREE(inp, size); return -error; } int venus_remove(struct super_block *sb, struct ViceFid *dirfid, const char *name, int length) { struct inputArgs *inp; struct outputArgs *outp; int error=0, size, payload_offset; payload_offset = VC_INSIZE(cfs_remove_in); size = CFS_MAXNAMLEN + payload_offset; UPARG(size, CFS_REMOVE); inp->d.cfs_remove.VFid = *dirfid; inp->d.cfs_remove.name = (char *)payload_offset; memcpy((char *)inp + payload_offset, name, size); error = coda_upcall(coda_sbp(sb), size, &size, inp); if ( inp ) CODA_FREE(inp, size); return -error; } int venus_readlink(struct super_block *sb, struct ViceFid *fid, char *buffer, int *length) { int error, size, retlen; char *result; struct inputArgs *inp; struct outputArgs *outp; char *buf=NULL; /*[CFS_MAXNAMLEN + VC_INSIZE(cfs_readlink_in)];*/ size = CFS_MAXPATHLEN + VC_INSIZE(cfs_readlink_in); UPARG(size, CFS_READLINK); inp->d.cfs_readlink.VFid = *fid; error = coda_upcall(coda_sbp(sb), size, &size, inp); if (! error) { retlen = outp->d.cfs_readlink.count; if ( retlen > *length ) retlen = *length; *length = retlen; result = (char *)outp + (int)outp->d.cfs_readlink.data; memcpy(buffer, result, retlen); } if (inp) CODA_FREE(buf, size); CDEBUG(D_INODE, " result %d\n",error); EXIT; return -error; } int venus_link(struct super_block *sb, struct ViceFid *fid, struct ViceFid *dirfid, const char *name, int len ) { int error, payload_offset, size; struct inputArgs *inp; struct outputArgs *outp; size = CFS_MAXNAMLEN + sizeof(struct inputArgs); UPARG(size, CFS_LINK); payload_offset = (VC_INSIZE(cfs_link_in)); inp->d.cfs_link.sourceFid = *fid; inp->d.cfs_link.destFid = *dirfid; inp->d.cfs_link.tname = (char *)payload_offset; /* make sure strings are null terminated */ memcpy((char *)inp + payload_offset, name, len); *((char *)inp + payload_offset + len) = '\0'; size = payload_offset + len + 1; error = coda_upcall(coda_sbp(sb), size, &size, inp); if (inp) CODA_FREE(inp, size); CDEBUG(D_INODE, " result %d\n",error); EXIT; return -error; } int venus_symlink(struct super_block *sb, struct ViceFid *fid, const char *name, int len, const char *symname, int symlen) { int error, payload_offset, size, s; struct inputArgs *inp; struct outputArgs *outp; /* * allocate space for regular input, * plus 1 path and 1 name, plus padding */ size = sizeof(struct inputArgs) + CFS_MAXNAMLEN + CFS_MAXNAMLEN + 8; UPARG(size, CFS_SYMLINK); /* inp->d.cfs_symlink.attr = *tva; XXXXXX */ inp->d.cfs_symlink.VFid = *fid; payload_offset = VC_INSIZE(cfs_symlink_in); inp->d.cfs_symlink.srcname =(char*) payload_offset; s = ( symlen & ~0x3 ) + 4; /* Round up to word boundary. */ /* don't forget to copy out the null termination */ memcpy((char *)inp + payload_offset, symname, symlen); *((char *)inp + payload_offset + symlen) = '\0'; payload_offset += s; inp->d.cfs_symlink.tname = (char *) payload_offset; s = (len & ~0x3) + 4; /* Round up to word boundary. */ memcpy((char *)inp + payload_offset, name, len); *((char *)inp + payload_offset + len) = '\0'; size = payload_offset + s; error = coda_upcall(coda_sbp(sb), size, &size, inp); if (inp) CODA_FREE(inp, size); CDEBUG(D_INODE, " result %d\n",error); EXIT; return -error; } int venus_access(struct super_block *sb, struct ViceFid *fid, int mask) { struct inputArgs *inp; struct outputArgs *outp; int size; int error; size = sizeof(struct inputArgs); UPARG(size, CFS_ACCESS); inp->d.cfs_access.VFid = *fid; inp->d.cfs_access.flags = mask << 6; error = coda_upcall(coda_sbp(sb), size, &size, inp); if (inp) CODA_FREE(inp, sizeof(struct inputArgs)); EXIT; return -error; } int venus_pioctl(struct super_block *sb, struct ViceFid *fid, unsigned int cmd, struct PioctlData *data) { struct inputArgs *inp; struct outputArgs *outp; int size, error = 0; int iocsize; char str[50]; size = VC_MAXMSGSIZE; UPARG(size, CFS_IOCTL); /* build packet for Venus */ if (data->vi.in_size > VC_DATASIZE) { error = EINVAL; goto exit; } inp->d.cfs_ioctl.VFid = *fid; /* the cmd field was mutated by increasing its size field to * reflect the path and follow args. We need to subtract that * out before sending the command to Venus. */ inp->d.cfs_ioctl.cmd = (cmd & ~(IOCPARM_MASK << 16)); iocsize = ((cmd >> 16) & IOCPARM_MASK) - sizeof(char *) - sizeof(int); inp->d.cfs_ioctl.cmd |= (iocsize & IOCPARM_MASK) << 16; /* in->d.cfs_ioctl.rwflag = flag; */ inp->d.cfs_ioctl.len = data->vi.in_size; inp->d.cfs_ioctl.data = (char *)(VC_INSIZE(cfs_ioctl_in)); /* get the data out of user space */ if ( copy_from_user((char*)inp + (int)inp->d.cfs_ioctl.data, data->vi.in, data->vi.in_size) ) { error = EINVAL; goto exit; } error = coda_upcall(coda_sbp(sb), size, &size, inp); if (error) { printk("coda_pioctl: Venus returns: %d for %s\n", error, coda_f2s(fid, str)); goto exit; } /* Copy out the OUT buffer. */ if (outp->d.cfs_ioctl.len > data->vi.out_size) { CDEBUG(D_FILE, "return len %d <= request len %d\n", outp->d.cfs_ioctl.len, data->vi.out_size); error = EINVAL; } else { if (copy_to_user(data->vi.out, (char *)outp + (int)outp->d.cfs_ioctl.data, data->vi.out_size)) { error = EINVAL; goto exit; } } exit: if (inp) CODA_FREE(inp, VC_MAXMSGSIZE); return -error; } /* * coda_upcall and coda_downcall routines. * */ /* * coda_upcall will return a POSITIVE error in the case of * failed communication with Venus _or_ will peek at Venus * reply and return Venus' error, also POSITIVE. * */ int coda_upcall(struct coda_sb_info *sbi, int inSize, int *outSize, struct inputArgs *buffer) { struct vcomm *vcommp; struct outputArgs *out; struct vmsg *vmp; int error = 0; ENTRY; if (sbi->sbi_vcomm == NULL) { return ENODEV; } vcommp = sbi->sbi_vcomm; clstats(((struct inputArgs *)buffer)->opcode); if (!vcomm_open(vcommp)) return(ENODEV); /* Format the request message. */ CODA_ALLOC(vmp,struct vmsg *,sizeof(struct vmsg)); vmp->vm_data = (void *)buffer; vmp->vm_flags = 0; vmp->vm_inSize = inSize; vmp->vm_outSize = *outSize ? *outSize : inSize; vmp->vm_opcode = ((struct inputArgs *)buffer)->opcode; vmp->vm_unique = ++vcommp->vc_seq; vmp->vm_sleep = NULL; /* Fill in the common input args. */ ((struct inputArgs *)buffer)->unique = vmp->vm_unique; /* Append msg to pending queue and poke Venus. */ INSQUE(vmp->vm_chain, vcommp->vc_pending); CDEBUG(D_UPCALL, "Proc %d wake Venus for(opc,uniq) =(%d,%d) msg at %x.zzz.\n", current->pid, vmp->vm_opcode, vmp->vm_unique, (int)vmp); wake_up_interruptible(&vcommp->vc_waitq); /* We can be interrupted while we wait for Venus to process * our request. If the interrupt occurs before Venus has read * the request, we dequeue and return. If it occurs after the * read but before the reply, we dequeue, send a signal * message, and return. If it occurs after the reply we ignore * it. In no case do we want to restart the syscall. If it * was interrupted by a venus shutdown (vcclose), return * ENODEV. */ /* Ignore return, We have to check anyway */ interruptible_sleep_on(&vmp->vm_sleep); CDEBUG(D_UPCALL, "..process %d woken up by Venus for vmp at 0x%x, data at %x\n", current->pid, (int)vmp, (int)vmp->vm_data); if (vcomm_open(vcommp)) { /* i.e. Venus is still alive */ /* Op went through, interrupt or not... */ if (vmp->vm_flags & VM_WRITE) { error = 0; out = (struct outputArgs *)vmp->vm_data; error = out->result; CDEBUG(D_UPCALL, "upcall: (u,o,r) (%ld, %ld, %ld) out at %x\n", out->unique, out->opcode, out->result, (int)out); *outSize = vmp->vm_outSize; goto exit; } if (!(vmp->vm_flags & VM_READ)) { /* Interrupted before venus read it. */ CDEBUG(D_UPCALL, "Interrupted before read:(op,un) (%d.%d), flags = %x\n", vmp->vm_opcode, vmp->vm_unique, vmp->vm_flags); REMQUE(vmp->vm_chain); error = ERESTARTSYS; goto exit; } if ( vmp->vm_flags & VM_READ) { /* interrupted after Venus did its read, send signal */ struct inputArgs *dog; struct vmsg *svmp; CDEBUG(D_UPCALL, "Sending Venus a signal: op = %d.%d, flags = %x\n", vmp->vm_opcode, vmp->vm_unique, vmp->vm_flags); REMQUE(vmp->vm_chain); error = ERESTARTSYS; CODA_ALLOC(svmp, struct vmsg *, sizeof (struct vmsg)); CODA_ALLOC((svmp->vm_data), char *, VC_IN_NO_DATA); CDEBUG(D_UPCALL, "coda_upcall: enqueing signal msg (%d, %d)\n", svmp->vm_opcode, svmp->vm_unique); dog = (struct inputArgs *)svmp->vm_data; dog->opcode = CFS_SIGNAL; dog->unique = vmp->vm_unique; svmp->vm_flags = 0; svmp->vm_opcode = dog->opcode; svmp->vm_unique = dog->unique; svmp->vm_inSize = VC_IN_NO_DATA; svmp->vm_outSize = VC_IN_NO_DATA; /* insert at head of queue! */ INSQUE(svmp->vm_chain, vcommp->vc_pending); wake_up_interruptible(&vcommp->vc_waitq); } } else { /* If venus died i.e. !VC_OPEN(vcommp) */ printk("coda_upcall: Venus dead upon (op,un) (%d.%d) flags %d\n", vmp->vm_opcode, vmp->vm_unique, vmp->vm_flags); /* if (! (vmp->vm_flags & VM_WRITE) ) */ error = ENODEV; } exit: CODA_FREE(vmp, sizeof(struct vmsg)); if (error) badclstats(); return error; } /* * There are 7 cases where cache invalidations occur. The semantics * of each is listed here: * * CFS_FLUSH -- flush all entries from the name cache and the cnode cache. * CFS_PURGEUSER -- flush all entries from the name cache for a specific user * This call is a result of token expiration. * Linux does a cfsnc_flush since cred's are not maintained. * * The next arise as the result of callbacks on a file or directory. * CFS_ZAPDIR -- flush the attributes for the dir from its cnode. * Zap all children of this directory from the namecache. * CFS_ZAPFILE -- flush the cached attributes for a file. * CFS_ZAPVNODE -- in linux the same as zap file (no creds). * * The next is a result of Venus detecting an inconsistent file. * CFS_PURGEFID -- flush the attribute for the file * If it is a dir (odd vnode), purge its * children from the namecache * remove the file from the namecache. * * The last allows Venus to replace local fids with global ones * during reintegration. * * CFS_REPLACE -- replace one ViceFid with another throughout the name cache */ int coda_downcall(int opcode, struct outputArgs * out) { /* Handle invalidate requests. */ switch (opcode) { case CFS_FLUSH : { clstats(CFS_FLUSH); cfsnc_flush(); return(0); } case CFS_PURGEUSER : { clstats(CFS_PURGEUSER); cfsnc_flush(); return(0); } case CFS_ZAPDIR : { ViceFid *fid = &out->d.cfs_zapdir.CodaFid; clstats(CFS_ZAPDIR); cfsnc_zapfid(fid); cfsnc_zapParentfid(fid); CDEBUG(D_UPCALL, "zapdir: fid = (%lx.%lx.%lx), \n",fid->Volume, fid->Vnode, fid->Unique); return(0); } case CFS_ZAPVNODE : { clstats(CFS_ZAPVNODE); cfsnc_zapfid(&out->d.cfs_zapvnode.VFid); return(0); } case CFS_ZAPFILE : { clstats(CFS_ZAPFILE); cfsnc_zapfid(&out->d.cfs_zapfile.CodaFid); return 0; } case CFS_PURGEFID : { ViceFid *fid = &out->d.cfs_purgefid.CodaFid; clstats(CFS_PURGEFID); cfsnc_zapfid(fid); cfsnc_zapParentfid(fid); CDEBUG(D_UPCALL, "purgefid: fid = (%lx.%lx.%lx)\n", fid->Volume, fid->Vnode, fid->Unique); return 0; } case CFS_REPLACE : { clstats(CFS_REPLACE); cfsnc_replace(&out->d.cfs_replace.OldFid, &out->d.cfs_replace.NewFid); return (0); } } return 0; }