| Message ID | 20200917082236.2518236-3-hch@lst.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/5] nfs: simplify nfs4_parse_monolithic | expand |
On Thu, Sep 17, 2020 at 10:22:33AM +0200, Christoph Hellwig wrote: > There is no reason the generic fs code should bother with NFS specific > binary mount data - lift the conversion into nfs4_parse_monolithic > instead. Considering the size of struct compat_nfs4_mount_data_v1... Do we really need to bother with that "copy in place, so we go through the fields backwards" logics? Just make that > +static void nfs4_compat_mount_data_conv(struct nfs4_mount_data *data) > +{ struct compat_nfs4_mount_data_v1 compat; compat = *(struct compat_nfs4_mount_data_v1 *)data; and copy the damnt thing without worrying about the field order...
On Thu, Sep 17, 2020 at 06:16:04PM +0100, Al Viro wrote: > On Thu, Sep 17, 2020 at 10:22:33AM +0200, Christoph Hellwig wrote: > > There is no reason the generic fs code should bother with NFS specific > > binary mount data - lift the conversion into nfs4_parse_monolithic > > instead. > > Considering the size of struct compat_nfs4_mount_data_v1... Do we really > need to bother with that "copy in place, so we go through the fields > backwards" logics? Just make that > > > +static void nfs4_compat_mount_data_conv(struct nfs4_mount_data *data) > > +{ > struct compat_nfs4_mount_data_v1 compat; > compat = *(struct compat_nfs4_mount_data_v1 *)data; > and copy the damnt thing without worrying about the field order... Maybe. But then again why bother? I just sticked to the existing code as much as possible.
On Thu, Sep 17, 2020 at 07:18:26PM +0200, Christoph Hellwig wrote: > On Thu, Sep 17, 2020 at 06:16:04PM +0100, Al Viro wrote: > > On Thu, Sep 17, 2020 at 10:22:33AM +0200, Christoph Hellwig wrote: > > > There is no reason the generic fs code should bother with NFS specific > > > binary mount data - lift the conversion into nfs4_parse_monolithic > > > instead. > > > > Considering the size of struct compat_nfs4_mount_data_v1... Do we really > > need to bother with that "copy in place, so we go through the fields > > backwards" logics? Just make that > > > > > +static void nfs4_compat_mount_data_conv(struct nfs4_mount_data *data) > > > +{ > > struct compat_nfs4_mount_data_v1 compat; > > compat = *(struct compat_nfs4_mount_data_v1 *)data; > > and copy the damnt thing without worrying about the field order... > > Maybe. But then again why bother? I just sticked to the existing > code as much as possible. Trond, Anna: what is your preference?
This is for the binary mount stuff? That was already legacy code when I first started, and mount uses text options now. My preference is for keeping it as close to the original code as possible. I'm curious if you've been able to test this? I'm not sure if there is a way to force binary mount data through mount.nfs Anna On Mon, Sep 21, 2020 at 2:49 AM Christoph Hellwig <hch@lst.de> wrote: > > On Thu, Sep 17, 2020 at 07:18:26PM +0200, Christoph Hellwig wrote: > > On Thu, Sep 17, 2020 at 06:16:04PM +0100, Al Viro wrote: > > > On Thu, Sep 17, 2020 at 10:22:33AM +0200, Christoph Hellwig wrote: > > > > There is no reason the generic fs code should bother with NFS specific > > > > binary mount data - lift the conversion into nfs4_parse_monolithic > > > > instead. > > > > > > Considering the size of struct compat_nfs4_mount_data_v1... Do we really > > > need to bother with that "copy in place, so we go through the fields > > > backwards" logics? Just make that > > > > > > > +static void nfs4_compat_mount_data_conv(struct nfs4_mount_data *data) > > > > +{ > > > struct compat_nfs4_mount_data_v1 compat; > > > compat = *(struct compat_nfs4_mount_data_v1 *)data; > > > and copy the damnt thing without worrying about the field order... > > > > Maybe. But then again why bother? I just sticked to the existing > > code as much as possible. > > Trond, Anna: what is your preference?
On Mon, Sep 21, 2020 at 12:05:52PM -0400, Anna Schumaker wrote: > This is for the binary mount stuff? That was already legacy code when > I first started, and mount uses text options now. My preference is for > keeping it as close to the original code as possible. Ok. Al, are you fine with the series as-is then? > > I'm curious if you've been able to test this? I'm not sure if there is > a way to force binary mount data through mount.nfs The test is pretty trivial - even the most recent nfsutils supports the binary mount data. You just need to patch out the detection for a modern kernel.
On Mon, Sep 21, 2020 at 08:11:23PM +0200, Christoph Hellwig wrote: > On Mon, Sep 21, 2020 at 12:05:52PM -0400, Anna Schumaker wrote: > > This is for the binary mount stuff? That was already legacy code when > > I first started, and mount uses text options now. My preference is for > > keeping it as close to the original code as possible. > > Ok. Al, are you fine with the series as-is then? I can live with that. I'm not fond of in_compat_syscall() proliferation, but in this case it's reasonably sane... OK, applied.
diff --git a/fs/compat.c b/fs/compat.c index 436d228cf71c09..9b00523d7fa571 100644 --- a/fs/compat.c +++ b/fs/compat.c @@ -19,73 +19,6 @@ #include <linux/uaccess.h> #include "internal.h" -struct compat_nfs_string { - compat_uint_t len; - compat_uptr_t data; -}; - -static inline void compat_nfs_string(struct nfs_string *dst, - struct compat_nfs_string *src) -{ - dst->data = compat_ptr(src->data); - dst->len = src->len; -} - -struct compat_nfs4_mount_data_v1 { - compat_int_t version; - compat_int_t flags; - compat_int_t rsize; - compat_int_t wsize; - compat_int_t timeo; - compat_int_t retrans; - compat_int_t acregmin; - compat_int_t acregmax; - compat_int_t acdirmin; - compat_int_t acdirmax; - struct compat_nfs_string client_addr; - struct compat_nfs_string mnt_path; - struct compat_nfs_string hostname; - compat_uint_t host_addrlen; - compat_uptr_t host_addr; - compat_int_t proto; - compat_int_t auth_flavourlen; - compat_uptr_t auth_flavours; -}; - -static int do_nfs4_super_data_conv(void *raw_data) -{ - int version = *(compat_uint_t *) raw_data; - - if (version == 1) { - struct compat_nfs4_mount_data_v1 *raw = raw_data; - struct nfs4_mount_data *real = raw_data; - - /* copy the fields backwards */ - real->auth_flavours = compat_ptr(raw->auth_flavours); - real->auth_flavourlen = raw->auth_flavourlen; - real->proto = raw->proto; - real->host_addr = compat_ptr(raw->host_addr); - real->host_addrlen = raw->host_addrlen; - compat_nfs_string(&real->hostname, &raw->hostname); - compat_nfs_string(&real->mnt_path, &raw->mnt_path); - compat_nfs_string(&real->client_addr, &raw->client_addr); - real->acdirmax = raw->acdirmax; - real->acdirmin = raw->acdirmin; - real->acregmax = raw->acregmax; - real->acregmin = raw->acregmin; - real->retrans = raw->retrans; - real->timeo = raw->timeo; - real->wsize = raw->wsize; - real->rsize = raw->rsize; - real->flags = raw->flags; - real->version = raw->version; - } - - return 0; -} - -#define NFS4_NAME "nfs4" - COMPAT_SYSCALL_DEFINE5(mount, const char __user *, dev_name, const char __user *, dir_name, const char __user *, type, compat_ulong_t, flags, @@ -111,14 +44,6 @@ COMPAT_SYSCALL_DEFINE5(mount, const char __user *, dev_name, if (IS_ERR(options)) goto out2; - if (kernel_type && options) { - if (!strcmp(kernel_type, NFS4_NAME)) { - retval = -EINVAL; - if (do_nfs4_super_data_conv(options)) - goto out3; - } - } - retval = do_mount(kernel_dev, dir_name, kernel_type, flags, options); out3: diff --git a/fs/nfs/fs_context.c b/fs/nfs/fs_context.c index cbf6a4ba5e5806..222afba70bc08e 100644 --- a/fs/nfs/fs_context.c +++ b/fs/nfs/fs_context.c @@ -1039,6 +1039,65 @@ static int nfs23_parse_monolithic(struct fs_context *fc, } #if IS_ENABLED(CONFIG_NFS_V4) +struct compat_nfs_string { + compat_uint_t len; + compat_uptr_t data; +}; + +static inline void compat_nfs_string(struct nfs_string *dst, + struct compat_nfs_string *src) +{ + dst->data = compat_ptr(src->data); + dst->len = src->len; +} + +struct compat_nfs4_mount_data_v1 { + compat_int_t version; + compat_int_t flags; + compat_int_t rsize; + compat_int_t wsize; + compat_int_t timeo; + compat_int_t retrans; + compat_int_t acregmin; + compat_int_t acregmax; + compat_int_t acdirmin; + compat_int_t acdirmax; + struct compat_nfs_string client_addr; + struct compat_nfs_string mnt_path; + struct compat_nfs_string hostname; + compat_uint_t host_addrlen; + compat_uptr_t host_addr; + compat_int_t proto; + compat_int_t auth_flavourlen; + compat_uptr_t auth_flavours; +}; + +static void nfs4_compat_mount_data_conv(struct nfs4_mount_data *data) +{ + struct compat_nfs4_mount_data_v1 *compat = + (struct compat_nfs4_mount_data_v1 *)data; + + /* copy the fields backwards */ + data->auth_flavours = compat_ptr(compat->auth_flavours); + data->auth_flavourlen = compat->auth_flavourlen; + data->proto = compat->proto; + data->host_addr = compat_ptr(compat->host_addr); + data->host_addrlen = compat->host_addrlen; + compat_nfs_string(&data->hostname, &compat->hostname); + compat_nfs_string(&data->mnt_path, &compat->mnt_path); + compat_nfs_string(&data->client_addr, &compat->client_addr); + data->acdirmax = compat->acdirmax; + data->acdirmin = compat->acdirmin; + data->acregmax = compat->acregmax; + data->acregmin = compat->acregmin; + data->retrans = compat->retrans; + data->timeo = compat->timeo; + data->wsize = compat->wsize; + data->rsize = compat->rsize; + data->flags = compat->flags; + data->version = compat->version; +} + /* * Validate NFSv4 mount options */ @@ -1061,6 +1120,9 @@ static int nfs4_parse_monolithic(struct fs_context *fc, if (data->version != 1) return generic_parse_monolithic(fc, data); + if (in_compat_syscall()) + nfs4_compat_mount_data_conv(data); + if (data->host_addrlen > sizeof(ctx->nfs_server.address)) goto out_no_address; if (data->host_addrlen == 0)
There is no reason the generic fs code should bother with NFS specific binary mount data - lift the conversion into nfs4_parse_monolithic instead. Signed-off-by: Christoph Hellwig <hch@lst.de> --- fs/compat.c | 75 --------------------------------------------- fs/nfs/fs_context.c | 62 +++++++++++++++++++++++++++++++++++++ 2 files changed, 62 insertions(+), 75 deletions(-)