- 23 Sep, 2019 1 commit
-
-
YueHaibing authored
Fix sparse warning: fs/nfsd/nfssvc.c:364:6: warning: symbol 'nfsd_reset_boot_verifier_locked' was not declared. Should it be static? Reported-by:
Hulk Robot <hulkci@huawei.com> Signed-off-by:
YueHaibing <yuehaibing@huawei.com> Signed-off-by:
J. Bruce Fields <bfields@redhat.com>
-
- 20 Sep, 2019 2 commits
-
-
NeilBrown authored
This original code in nfsd4_get_drc_mem() would hand out 30 slots (approximately NFSD_MAX_MEM_PER_SESSION bytes at slightly over 2K per slot) to each requesting client until it ran out of space, then it would possibly give one last client a reduced allocation, then fail the allocation. Since commit de766e57 ("nfsd: give out fewer session slots as limit approaches") the last 90 slots to be given to about 12 clients with quickly reducing slot counts (better than just 3 clients). This still seems unnecessarily hasty. A subsequent patch allows over-allocation so every client gets at least one slot, but that might be a bit restrictive. The requested number of nfsd threads is the best guide we have to the expected number of clients, so use that - if it is at least 8. 256 threads on a 256Meg machine - which is a lot for a tiny machine - would result in nfsd_drc_max_mem being 2Meg, so 8K (3 slots) would be available for the first client, and over 200 clients would get more than 1 slot. So I don't think this change will be too debilitating on poorly configured machines, though it does mean that a sensible configuration is a little more important. Signed-off-by:
NeilBrown <neilb@suse.de> Signed-off-by:
-
NeilBrown authored
Currently, if there are more clients than allowed for by the space allocation in set_max_drc(), we fail a SESSION_CREATE request with NFS4ERR_DELAY. This means that the client retries indefinitely, which isn't a user-friendly response. The RFC requires NFS4ERR_NOSPC, but that would at best result in a clean failure on the client, which is not much more friendly. The current space allocation is a best-guess and doesn't provide any guarantees, we could still run out of space when trying to allocate drc space. So fail more gracefully - always give out at least one slot. If all clients used all the space in all slots, we might start getting memory pressure, but that is possible anyway. So ensure 'num' is always at least 1, and remove the test for it being zero. Signed-off-by:
-
- 10 Sep, 2019 6 commits
-
-
Scott Mayhew authored
Version 2 upcalls will allow the nfsd to include a hash of the kerberos principal string in the Cld_Create upcall. If a principal is present in the svc_cred, then the hash will be included in the Cld_Create upcall. We attempt to use the svc_cred.cr_raw_principal (which is returned by gssproxy) first, and then fall back to using the svc_cred.cr_principal (which is returned by both gssproxy and rpc.svcgssd). Upon a subsequent restart, the hash will be returned in the Cld_Gracestart downcall and stored in the reclaim_str_hashtbl so it can be used when handling reclaim opens. Signed-off-by:
Scott Mayhew <smayhew@redhat.com> Signed-off-by:
-
Scott Mayhew authored
Add a "GetVersion" upcall to allow nfsd to determine the maximum upcall version that the nfsdcld userspace daemon supports. If the daemon responds with -EOPNOTSUPP, then we know it only supports v1. Signed-off-by:
-
Trond Myklebust authored
If multiple clients are writing to the same file, then due to the fact we share a single file descriptor between all NFSv3 clients writing to the file, we have a situation where clients can miss the fact that their file data was not persisted. While this should be rare, it could cause silent data loss in situations where multiple clients are using NLM locking or O_DIRECT to write to the same file. Unfortunately, the stateless nature of NFSv3 and the fact that we can only identify clients by their IP address means that we cannot trivially cache errors; we would not know when it is safe to release them from the cache. So the solution is to declare a reboot. We understand that this should be a rare occurrence, since disks are usually stable. The most frequent occurrence is likely to be ENOSPC, at which point all writes to the given filesystem are likely to fail anyway. So the expectation is that clients will be forced to retry their writes until they hit the fatal error. Signed-off-by:
Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by:
-
Trond Myklebust authored
If a file may contain unstable writes that can error out, then we want to avoid garbage collecting the struct nfsd_file that may be tracking those errors. So in the garbage collector, we try to avoid collecting files that aren't clean. Furthermore, we avoid immediately kicking off the garbage collector in the case where the reference drops to zero for the case where there is a write error that is being tracked. If the file is unhashed while an error is pending, then declare a reboot, to ensure the client resends any unstable writes. Signed-off-by:
-
Trond Myklebust authored
Add support to allow the server to reset the boot verifier in order to force clients to resend I/O after a timeout failure. Signed-off-by:
Lance Shelton <lance.shelton@hammerspace.com> Signed-off-by:
-
Trond Myklebust authored
Ensure that we can safely clear out the file cache entries when the nfs server is shut down on a container. Otherwise, the file cache may end up pinning the mounts. Signed-off-by:
-
- 29 Aug, 2019 1 commit
-
-
J. Bruce Fields authored
We're unnecessarily limiting the size of an ACL to less than what most filesystems will support. Some users do hit the limit and it's confusing and unnecessary. It still seems prudent to impose some limit on the number of ACEs the client gives us before passing it straight to kmalloc(). So, let's just limit it to the maximum number that would be possible given the amount of data left in the argument buffer. That will still leave one limit beyond whatever the filesystem imposes: the client and server negotiate a limit on the size of a request, which we have to respect. But we're no longer imposing any additional arbitrary limit. struct nfs4_ace is 20 bytes on my system and the maximum call size we'll negotiate is about a megabyte, so in practice this is limiting the allocation here to about a megabyte. Reported-by:
"de Vandiere, Louis" <louis.devandiere@atos.net> Signed-off-by:
-
- 26 Aug, 2019 1 commit
-
-
J. Bruce Fields authored
This is only useful for client testing. I haven't really maintained it, and reference counting and locking are wrong at this point. You can get some of the same functionality now from nfsd/clients/. It was a good idea but I think its time has passed. In the unlikely event of users, hopefully the BROKEN dependency will prompt them to speak up. Otherwise I expect to remove it soon. Reported-by:
Alex Lyakas <alex@zadara.com> Signed-off-by:
-
- 20 Aug, 2019 1 commit
-
-
YueHaibing authored
Remove duplicated include. Signed-off-by:
-
- 19 Aug, 2019 18 commits
-
-
Trond Myklebust authored
Signed-off-by:
-
Trond Myklebust authored
Signed-off-by:
-
Jeff Layton authored
It's not uncommon for some workloads to do a bunch of I/O to a file and delete it just afterward. If knfsd has a cached open file however, then the file may still be open when the dentry is unlinked. If the underlying filesystem is nfs, then that could trigger it to do a sillyrename. On a REMOVE or RENAME scan the nfsd_file cache for open files that correspond to the inode, and proactively unhash and put their references. This should prevent any delete-on-last-close activity from occurring, solely due to knfsd's open file cache. This must be done synchronously though so we use the variants that call flush_delayed_fput. There are deadlock possibilities if you call flush_delayed_fput while holding locks, however. In the case of nfsd_rename, we don't even do the lookups of the dentries to be renamed until we've locked for rename. Once we've figured out what the target dentry is for a rename, check to see whether there are cached open files associated with it. If there are, then unwind all of the locking, close them all, and then reattempt the rename. Signed-off-by:
Jeff Layton <jeff.layton@primarydata.com> Signed-off-by:
Trond Myklebust <trond.myklebust@primarydata.com> Signed-off-by:
-
Jeff Layton authored
The raparms cache was set up in order to ensure that we carry readahead information forward from one RPC call to the next. In other words, it was set up because each RPC call was forced to open a struct file, then close it, causing the loss of readahead information that is normally cached in that struct file, and used to keep the page cache filled when a user calls read() multiple times on the same file descriptor. Now that we cache the struct file, and reuse it for all the I/O calls to a given file by a given user, we no longer have to keep a separate readahead cache. Signed-off-by:
-
Jeff Layton authored
Signed-off-by:
-
Jeff Layton authored
Have nfs4_preprocess_stateid_op pass back a nfsd_file instead of a filp. Since we now presume that the struct file will be persistent in most cases, we can stop fiddling with the raparms in the read code. This also means that we don't really care about the rd_tmp_file field anymore. Signed-off-by:
-
Jeff Layton authored
Have them keep an nfsd_file reference instead of a struct file. Signed-off-by:
-
Jeff Layton authored
Signed-off-by:
-
Jeff Layton authored
Use cached filps if possible instead of opening a new one every time. Signed-off-by:
-
Jeff Layton authored
Signed-off-by:
-
Jeff Layton authored
Signed-off-by:
-
Jeff Layton authored
Currently, NFSv2/3 reads and writes have to open a file, do the read or write and then close it again for each RPC. This is highly inefficient, especially when the underlying filesystem has a relatively slow open routine. This patch adds a new open file cache to knfsd. Rather than doing an open for each RPC, the read/write handlers can call into this cache to see if there is one already there for the correct filehandle and NFS_MAY_READ/WRITE flags. If there isn't an entry, then we create a new one and attempt to perform the open. If there is, then we wait until the entry is fully instantiated and return it if it is at the end of the wait. If it's not, then we attempt to take over construction. Since the main goal is to speed up NFSv2/3 I/O, we don't want to close these files on last put of these objects. We need to keep them around for a little while since we never know when the next READ/WRITE will come in. Cache entries have a hardcoded 1s timeout, and we have a recurring workqueue job that walks the cache and purges any entries that have expired. Signed-off-by:
Weston Andros Adamson <dros@primarydata.com> Signed-off-by:
Richard Sharpe <richard.sharpe@primarydata.com> Signed-off-by:
-
Trond Myklebust authored
Allow knfsd to flush the delayed fput list so that it can ensure the cached struct file is closed before it is unlinked. Signed-off-by:
-
Trond Myklebust authored
The knfsd file cache will need to detect when files are unlinked, so that it can close the associated cached files. Export a minimal set of notifier functions to allow it to do so. Signed-off-by:
-
Jeff Layton authored
With the new file caching infrastructure in nfsd, we can end up holding files open for an indefinite period of time, even when they are still idle. This may prevent the kernel from handing out leases on the file, which is something we don't want to block. Fix this by running a SRCU notifier call chain whenever on any lease attempt. nfsd can then purge the cache for that inode before returning. Since SRCU is only conditionally compiled in, we must only define the new chain if it's enabled, and users of the chain must ensure that SRCU is enabled. Signed-off-by:
-
Jeff Layton authored
When the exports table is changed, exportfs will usually write a new time to the "flush" file in the nfsd.export cache procfile. This tells the kernel to flush any entries that are older than that value. This gives us a mechanism to tell whether an unexport might have occurred. Add a new ->flush cache_detail operation that is called after flushing the cache whenever someone writes to a "flush" file. Signed-off-by:
-
Chuck Lever authored
Use a wait-free mechanism for managing the svc_rdma_recv_ctxts free list. Subsequently, sc_recv_lock can be eliminated. Signed-off-by:
Chuck Lever <chuck.lever@oracle.com> Signed-off-by:
-
Chuck Lever authored
Clean up: the system workqueue will work just as well. Signed-off-by:
-
- 16 Aug, 2019 3 commits
-
-
J. Bruce Fields authored
"cb" is never actually NULL in these functions. On a quick skim of the history, they seem to have been there from the beginning. I'm not sure if they originally served a purpose. Reported-by:
Jia-Ju Bai <baijiaju1990@gmail.com> Signed-off-by:
-
J. Bruce Fields authored
These are some nfsd bugfixes that I also want in the main branch that will be submitted for the 5.4 merge window.
-
He Zhe authored
reply_cache_stats uses wrong parameter as seq file private structure and thus causes the following kernel crash when users read /proc/fs/nfsd/reply_cache_stats BUG: kernel NULL pointer dereference, address: 00000000000001f9 PGD 0 P4D 0 Oops: 0000 [#3] SMP PTI CPU: 6 PID: 1502 Comm: cat Tainted: G D 5.3.0-rc3+ #1 Hardware name: Intel Corporation Broadwell Client platform/Basking Ridge, BIOS BDW-E2R1.86C.0118.R01.1503110618 03/11/2015 RIP: 0010:nfsd_reply_cache_stats_show+0x3b/0x2d0 Code: 41 54 49 89 f4 48 89 fe 48 c7 c7 b3 10 33 88 53 bb e8 03 00 00 e8 88 82 d1 ff bf 58 89 41 00 e8 eb c5 85 00 48 83 eb 01 75 f0 <41> 8b 94 24 f8 01 00 00 48 c7 c6 be 10 33 88 4c 89 ef bb e8 03 00 RSP: 0018:ffffaa520106fe08 EFLAGS: 00010246 RAX: 000000cfe1a77123 RBX: 0000000000000000 RCX: 0000000000291b46 RDX: 000000cf00000000 RSI: 0000000000000006 RDI: 0000000000291b28 RBP: ffffaa520106fe20 R08: 0000000000000006 R09: 000000cfe17e55dd R10: ffffa424e47c0000 R11: 000000000000030b R12: 0000000000000001 R13: ffffa424e5697000 R14: 0000000000000001 R15: ffffa424e5697000 FS: 00007f805735f580(0000) GS:ffffa424f8f80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000000001f9 CR3: 00000000655ce005 CR4: 00000000003606e0 Call Trace: seq_read+0x194/0x3e0 __vfs_read+0x1b/0x40 vfs_read+0x95/0x140 ksys_read+0x61/0xe0 __x64_sys_read+0x1a/0x20 do_syscall_64+0x4d/0x120 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x7f805728b861 Code: fe ff ff 50 48 8d 3d 86 b4 09 00 e8 79 e0 01 00 66 0f 1f 84 00 00 00 00 00 48 8d 05 d9 19 0d 00 8b 00 85 c0 75 13 31 c0 0f 05 <48> 3d 00 f0 ff ff 77 57 c3 66 0f 1f 44 00 00 48 83 ec 28 48 89 54 RSP: 002b:00007ffea1ce3c38 EFLAGS: 00000246 ORIG_RAX: 0000000000000000 RAX: ffffffffffffffda RBX: 0000000000020000 RCX: 00007f805728b861 RDX: 0000000000020000 RSI: 00007f8057183000 RDI: 0000000000000003 RBP: 00007f8057183000 R08: 00007f8057182010 R09: 0000000000000000 R10: 0000000000000022 R11: 0000000000000246 R12: 0000559a60e8ff10 R13: 0000000000000003 R14: 0000000000020000 R15: 0000000000020000 Modules linked in: CR2: 00000000000001f9 ---[ end trace 01613595153f0cba ]--- RIP: 0010:nfsd_reply_cache_stats_show+0x3b/0x2d0 Code: 41 54 49 89 f4 48 89 fe 48 c7 c7 b3 10 33 88 53 bb e8 03 00 00 e8 88 82 d1 ff bf 58 89 41 00 e8 eb c5 85 00 48 83 eb 01 75 f0 <41> 8b 94 24 f8 01 00 00 48 c7 c6 be 10 33 88 4c 89 ef bb e8 03 00 RSP: 0018:ffffaa52004b3e08 EFLAGS: 00010246 RAX: 0000002bab45a7c6 RBX: 0000000000000000 RCX: 0000000000291b4c RDX: 0000002b00000000 RSI: 0000000000000004 RDI: 0000000000291b28 RBP: ffffaa52004b3e20 R08: 0000000000000004 R09: 0000002bab1c8c7a R10: ffffa424e5500000 R11: 00000000000002a9 R12: 0000000000000001 R13: ffffa424e4475000 R14: 0000000000000001 R15: ffffa424e4475000 FS: 00007f805735f580(0000) GS:ffffa424f8f80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000000001f9 CR3: 00000000655ce005 CR4: 00000000003606e0 Killed Fixes: 3ba75830 ("nfsd4: drc containerization") Signed-off-by:
He Zhe <zhe.he@windriver.com> Signed-off-by:
-
- 15 Aug, 2019 3 commits
-
-
J. Bruce Fields authored
A process could race in an open and attempt to read one of these files before i_private is initialized, and get a spurious error. Reported-by:
Al Viro <viro@zeniv.linux.org.uk> Signed-off-by:
-
J. Bruce Fields authored
synchronize_rcu() gets called multiple times each time a client is destroyed. If the laundromat thread has a lot of clients to destroy, the delay can be noticeable. This was causing pynfs test RENEW3 to fail. We could embed an rcu_head in each inode and do the kref_put in an rcu callback. But simplest is just to take a lock here. (I also wonder if the laundromat thread would be better replaced by a bunch of scheduled work or timers or something.) Signed-off-by: -
Tetsuo Handa authored
syzbot is reporting that nfsd_mkdir() forgot to remove dentry created by d_alloc_name() when __nfsd_mkdir() failed (due to memory allocation fault injection) [1]. [1] https://syzkaller.appspot.com/bug?id=ce41a1f769ea4637ebffedf004a803e8405b4674Signed-off-by:
Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Reported-by:
syzbot <syzbot+2c95195d5d433f6ed6cb@syzkaller.appspotmail.com> Fixes: e8a79fb1 ("nfsd: add nfsd/clients directory") [bfields: clean up in nfsd_mkdir instead of __nfsd_mkdir] Signed-off-by:
-
- 30 Jul, 2019 1 commit
-
-
Dave Wysochanski authored
The sunrpc cache interface is susceptible to being fooled by a rogue process just reading a 'channel' file. If this happens the kernel may think a valid daemon exists to service the cache when it does not. For example, the following may fool the kernel: cat /proc/net/rpc/auth.unix.gid/channel Change the tracking of readers to writers when considering whether a listener exists as all valid daemon processes either open a channel file O_RDWR or O_WRONLY. While this does not prevent a rogue process from "stealing" a message from the kernel, it does at least improve the kernels perception of whether a valid process servicing the cache exists. Signed-off-by:
Dave Wysochanski <dwysocha@redhat.com> Signed-off-by:
-
- 28 Jul, 2019 3 commits
-
-
Linus Torvalds authored
-
git://git.kernel.org/pub/scm/linux/kernel/git/kees/linuxLinus Torvalds authored
Pull structleak fix from Kees Cook: "Disable gcc-based stack variable auto-init under KASAN (Arnd Bergmann). This fixes a bunch of build warnings under KASAN and the gcc-plugin-based stack auto-initialization features (which are arguably redundant, so better to let KASAN control this)" * tag 'meminit-v5.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux: structleak: disable STRUCTLEAK_BYREF in combination with KASAN_STACK
-
Linus Torvalds authored
Merge tag 'kbuild-fixes-v5.3' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild Pull Kbuild fixes from Masahiro Yamada: - add compile_commands.json to .gitignore - fix false-positive warning from gen_compile_commands.py after allnoconfig build - remove unused code * tag 'kbuild-fixes-v5.3' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild: kbuild: remove unused single-used-m gen_compile_commands: lower the entry count threshold .gitignore: Add compilation database file kbuild: remove unused objectify macro
-
