mirror of
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git
synced 2024-09-04 01:49:36 +00:00
![Jeff Layton](/assets/img/avatar_default.png)
We currently can hit a deadlock (of sorts) when trying to use flexfiles layouts with XFS. XFS will call break_layout when something wants to write to the file. In the case of the (super-simple) flexfiles layout driver in knfsd, the MDS and DS are the same machine. The client can get a layout and then issue a v3 write to do its I/O. XFS will then call xfs_break_layouts, which will cause a CB_LAYOUTRECALL to be issued to the client. The client however can't return the layout until the v3 WRITE completes, but XFS won't allow the write to proceed until the layout is returned. Christoph says: XFS only cares about block-like layouts where the client has direct access to the file blocks. I'd need to look how to propagate the flag into break_layout, but in principle we don't need to do any recalls on truncate ever for file and flexfile layouts. If we're never going to recall the layout, then we don't even need to set the lease at all. Just skip doing so on flexfiles layouts by adding a new flag to struct nfsd4_layout_ops and skipping the lease setting and removal when that flag is true. Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
134 lines
3.4 KiB
C
134 lines
3.4 KiB
C
/*
|
|
* Copyright (c) 2016 Tom Haynes <loghyr@primarydata.com>
|
|
*
|
|
* The following implements a super-simple flex-file server
|
|
* where the NFSv4.1 mds is also the ds. And the storage is
|
|
* the same. I.e., writing to the mds via a NFSv4.1 WRITE
|
|
* goes to the same location as the NFSv3 WRITE.
|
|
*/
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/nfsd/debug.h>
|
|
|
|
#include <linux/sunrpc/addr.h>
|
|
|
|
#include "flexfilelayoutxdr.h"
|
|
#include "pnfs.h"
|
|
|
|
#define NFSDDBG_FACILITY NFSDDBG_PNFS
|
|
|
|
static __be32
|
|
nfsd4_ff_proc_layoutget(struct inode *inode, const struct svc_fh *fhp,
|
|
struct nfsd4_layoutget *args)
|
|
{
|
|
struct nfsd4_layout_seg *seg = &args->lg_seg;
|
|
u32 device_generation = 0;
|
|
int error;
|
|
uid_t u;
|
|
|
|
struct pnfs_ff_layout *fl;
|
|
|
|
/*
|
|
* The super simple flex file server has 1 mirror, 1 data server,
|
|
* and 1 file handle. So instead of 4 allocs, do 1 for now.
|
|
* Zero it out for the stateid - don't want junk in there!
|
|
*/
|
|
error = -ENOMEM;
|
|
fl = kzalloc(sizeof(*fl), GFP_KERNEL);
|
|
if (!fl)
|
|
goto out_error;
|
|
args->lg_content = fl;
|
|
|
|
/*
|
|
* Avoid layout commit, try to force the I/O to the DS,
|
|
* and for fun, cause all IOMODE_RW layout segments to
|
|
* effectively be WRITE only.
|
|
*/
|
|
fl->flags = FF_FLAGS_NO_LAYOUTCOMMIT | FF_FLAGS_NO_IO_THRU_MDS |
|
|
FF_FLAGS_NO_READ_IO;
|
|
|
|
/* Do not allow a IOMODE_READ segment to have write pemissions */
|
|
if (seg->iomode == IOMODE_READ) {
|
|
u = from_kuid(&init_user_ns, inode->i_uid) + 1;
|
|
fl->uid = make_kuid(&init_user_ns, u);
|
|
} else
|
|
fl->uid = inode->i_uid;
|
|
fl->gid = inode->i_gid;
|
|
|
|
error = nfsd4_set_deviceid(&fl->deviceid, fhp, device_generation);
|
|
if (error)
|
|
goto out_error;
|
|
|
|
fl->fh.size = fhp->fh_handle.fh_size;
|
|
memcpy(fl->fh.data, &fhp->fh_handle.fh_base, fl->fh.size);
|
|
|
|
/* Give whole file layout segments */
|
|
seg->offset = 0;
|
|
seg->length = NFS4_MAX_UINT64;
|
|
|
|
dprintk("GET: 0x%llx:0x%llx %d\n", seg->offset, seg->length,
|
|
seg->iomode);
|
|
return 0;
|
|
|
|
out_error:
|
|
seg->length = 0;
|
|
return nfserrno(error);
|
|
}
|
|
|
|
static __be32
|
|
nfsd4_ff_proc_getdeviceinfo(struct super_block *sb, struct svc_rqst *rqstp,
|
|
struct nfs4_client *clp, struct nfsd4_getdeviceinfo *gdp)
|
|
{
|
|
struct pnfs_ff_device_addr *da;
|
|
|
|
u16 port;
|
|
char addr[INET6_ADDRSTRLEN];
|
|
|
|
da = kzalloc(sizeof(struct pnfs_ff_device_addr), GFP_KERNEL);
|
|
if (!da)
|
|
return nfserrno(-ENOMEM);
|
|
|
|
gdp->gd_device = da;
|
|
|
|
da->version = 3;
|
|
da->minor_version = 0;
|
|
|
|
da->rsize = svc_max_payload(rqstp);
|
|
da->wsize = da->rsize;
|
|
|
|
rpc_ntop((struct sockaddr *)&rqstp->rq_daddr,
|
|
addr, INET6_ADDRSTRLEN);
|
|
if (rqstp->rq_daddr.ss_family == AF_INET) {
|
|
struct sockaddr_in *sin;
|
|
|
|
sin = (struct sockaddr_in *)&rqstp->rq_daddr;
|
|
port = ntohs(sin->sin_port);
|
|
snprintf(da->netaddr.netid, FF_NETID_LEN + 1, "tcp");
|
|
da->netaddr.netid_len = 3;
|
|
} else {
|
|
struct sockaddr_in6 *sin6;
|
|
|
|
sin6 = (struct sockaddr_in6 *)&rqstp->rq_daddr;
|
|
port = ntohs(sin6->sin6_port);
|
|
snprintf(da->netaddr.netid, FF_NETID_LEN + 1, "tcp6");
|
|
da->netaddr.netid_len = 4;
|
|
}
|
|
|
|
da->netaddr.addr_len =
|
|
snprintf(da->netaddr.addr, FF_ADDR_LEN + 1,
|
|
"%s.%hhu.%hhu", addr, port >> 8, port & 0xff);
|
|
|
|
da->tightly_coupled = false;
|
|
|
|
return 0;
|
|
}
|
|
|
|
const struct nfsd4_layout_ops ff_layout_ops = {
|
|
.notify_types =
|
|
NOTIFY_DEVICEID4_DELETE | NOTIFY_DEVICEID4_CHANGE,
|
|
.disable_recalls = true,
|
|
.proc_getdeviceinfo = nfsd4_ff_proc_getdeviceinfo,
|
|
.encode_getdeviceinfo = nfsd4_ff_encode_getdeviceinfo,
|
|
.proc_layoutget = nfsd4_ff_proc_layoutget,
|
|
.encode_layoutget = nfsd4_ff_encode_layoutget,
|
|
};
|