| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <m18vra7rzs.fsf@fess.ebiederm.org>
Date: Wed, 03 Aug 2011 02:40:23 -0700
From: ebiederm@...ssion.com (Eric W. Biederman)
To: Lucas De Marchi <lucas.demarchi@...fusion.mobi>
Cc: Greg KH <gregkh@...e.de>, Kay Sievers <kay.sievers@...y.org>,
Andrew Morton <akpm@...ux-foundation.org>,
Alan Cox <alan@...rguk.ukuu.org.uk>,
linux-kernel@...r.kernel.org, Nick Piggin <npiggin@...nel.dk>,
Al Viro <viro@...iv.linux.org.uk>,
Christoph Hellwig <hch@....de>,
Stephen Rothwell <sfr@...b.auug.org.au>,
David Howells <dhowells@...hat.com>,
"Serge E. Hallyn" <serge.hallyn@...onical.com>,
Daniel Lezcano <daniel.lezcano@...e.fr>,
Jiri Slaby <jslaby@...e.cz>, James Morris <jmorris@...ei.org>,
neilb@...e.de
Subject: Re: [PATCH] sysctl: add support for poll()
Lucas De Marchi <lucas.demarchi@...fusion.mobi> writes:
>>>From dae691ba16a6c6c50ca05e2cd69bdc808f2c4f89 Mon Sep 17 00:00:00 2001
> From: Lucas De Marchi <lucas.demarchi@...fusion.mobi>
> Date: Wed, 1 Jun 2011 09:14:36 -0300
> Subject: [PATCH] sysctl: add support for poll()
> Cc: linux-kernel@...r.kernel.org
>
> Adding support for poll() in sysctl fs allows userspace to receive
> notifications when an entry in sysctl changes. This way it's possible to
> know when hostname/domainname is changed due to the respective syscall
> has been called or its file under /proc/sys has been written to.
The patch breaks the return code of poll, is ugly, and it increases the
size of struct ctl_table.
I'm pretty certain I mentioned the return code breakage before.
In principle this is reasonable with the POLLERR | POLLPRI idiom but
I'm not a fan of this code.
Eric
> Signed-off-by: Lucas De Marchi <lucas.demarchi@...fusion.mobi>
> Acked-By: Kay Sievers <kay.sievers@...y.org>
> ---
> fs/proc/proc_sysctl.c | 40 ++++++++++++++++++++++++++++++++++++++++
> include/linux/sysctl.h | 8 ++++++++
> include/linux/utsname.h | 16 ++++++++++++++++
> kernel/sys.c | 2 ++
> kernel/utsname_sysctl.c | 36 ++++++++++++++++++++++++++++++++++++
> 5 files changed, 102 insertions(+), 0 deletions(-)
>
> diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
> index 1a77dbe..cf4bc7c 100644
> --- a/fs/proc/proc_sysctl.c
> +++ b/fs/proc/proc_sysctl.c
> @@ -3,6 +3,7 @@
> */
> #include <linux/init.h>
> #include <linux/sysctl.h>
> +#include <linux/poll.h>
> #include <linux/proc_fs.h>
> #include <linux/security.h>
> #include <linux/namei.h>
> @@ -176,6 +177,43 @@ static ssize_t proc_sys_write(struct file *filp, const char __user *buf,
> return proc_sys_call_handler(filp, (void __user *)buf, count, ppos, 1);
> }
>
> +static int proc_sys_open(struct inode *inode, struct file *filp)
> +{
> + struct ctl_table *table = PROC_I(inode)->sysctl_entry;
> +
> + if (table->poll) {
> + unsigned long event = atomic_read(&table->poll->event);
> +
> + filp->private_data = (void *)event;
It would be nice if there was something that abstracted
filp->private_data accesses so they were type safe, and
clear what you were doing.
> + }
> +
> + return 0;
> +}
> +
> +static unsigned int proc_sys_poll(struct file *filp, poll_table *wait)
> +{
> + struct inode *inode = filp->f_path.dentry->d_inode;
> + struct ctl_table *table = PROC_I(inode)->sysctl_entry;
> + unsigned long event = (unsigned long)filp->private_data;
> + unsigned int ret = POLLIN | POLLRDNORM;
The default return value here must be DEFAULT_POLLMASK otherwise you
change the result of poll for every single proc file and in general
return the wrong values because most sysctls are writable.
> + if (!table->proc_handler)
> + goto out;
> +
> + if (!table->poll)
> + goto out;
> +
> + poll_wait(filp, &table->poll->wait, wait);
> +
> + if (event != atomic_read(&table->poll->event)) {
> + filp->private_data = (void *)(unsigned long)atomic_read(
> + &table->poll->event);
> + ret = POLLIN | POLLRDNORM | POLLERR | POLLPRI;
Why are you updating the event here instead of when the file is read?
Presumably we want until the file is read and people see the new value.
> + }
> +
> +out:
> + return ret;
> +}
>
> static int proc_sys_fill_cache(struct file *filp, void *dirent,
> filldir_t filldir,
> @@ -364,6 +402,8 @@ static int proc_sys_getattr(struct vfsmount *mnt, struct dentry *dentry, struct
> }
>
> static const struct file_operations proc_sys_file_operations = {
> + .open = proc_sys_open,
> + .poll = proc_sys_poll,
> .read = proc_sys_read,
> .write = proc_sys_write,
> .llseek = default_llseek,
> diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
> index 11684d9..96c89ba 100644
> --- a/include/linux/sysctl.h
> +++ b/include/linux/sysctl.h
> @@ -25,6 +25,7 @@
> #include <linux/kernel.h>
> #include <linux/types.h>
> #include <linux/compiler.h>
> +#include <linux/wait.h>
>
> struct completion;
>
> @@ -1011,6 +1012,12 @@ extern int proc_do_large_bitmap(struct ctl_table *, int,
> * cover common cases.
> */
>
> +/* Support for userspace poll() to watch for changes */
> +struct ctl_table_poll {
> + atomic_t event;
> + wait_queue_head_t wait;
> +};
It is a moderately big stretch to think that everything that wants to
poll a value under /proc/sys will want to use struct ctl_table_poll.
> /* A sysctl table is an array of struct ctl_table: */
> struct ctl_table
> {
> @@ -1021,6 +1028,7 @@ struct ctl_table
> struct ctl_table *child;
> struct ctl_table *parent; /* Automatically set */
> proc_handler *proc_handler; /* Callback for text formatting */
> + struct ctl_table_poll *poll;
> void *extra1;
> void *extra2;
> };
> diff --git a/include/linux/utsname.h b/include/linux/utsname.h
> index 4e5b021..c714ed7 100644
> --- a/include/linux/utsname.h
> +++ b/include/linux/utsname.h
> @@ -37,6 +37,14 @@ struct new_utsname {
> #include <linux/nsproxy.h>
> #include <linux/err.h>
>
> +enum uts_proc {
> + UTS_PROC_OSTYPE,
> + UTS_PROC_OSRELEASE,
> + UTS_PROC_VERSION,
> + UTS_PROC_HOSTNAME,
> + UTS_PROC_DOMAINNAME,
> +};
> +
> struct user_namespace;
> extern struct user_namespace init_user_ns;
>
> @@ -80,6 +88,14 @@ static inline struct uts_namespace *copy_utsname(unsigned long flags,
> }
> #endif
>
> +#ifdef CONFIG_PROC_SYSCTL
> +extern void uts_proc_notify(enum uts_proc proc);
> +#else
> +static inline void uts_proc_notify(enum uts_proc proc)
> +{
> +}
> +#endif
> +
> static inline struct new_utsname *utsname(void)
> {
> return ¤t->nsproxy->uts_ns->name;
> diff --git a/kernel/sys.c b/kernel/sys.c
> index a101ba3..2347043 100644
> --- a/kernel/sys.c
> +++ b/kernel/sys.c
> @@ -1241,6 +1241,7 @@ SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
> memset(u->nodename + len, 0, sizeof(u->nodename) - len);
> errno = 0;
> }
> + uts_proc_notify(UTS_PROC_HOSTNAME);
> up_write(&uts_sem);
> return errno;
> }
> @@ -1291,6 +1292,7 @@ SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
> memset(u->domainname + len, 0, sizeof(u->domainname) - len);
> errno = 0;
> }
> + uts_proc_notify(UTS_PROC_DOMAINNAME);
> up_write(&uts_sem);
> return errno;
> }
> diff --git a/kernel/utsname_sysctl.c b/kernel/utsname_sysctl.c
> index a2cd77e..e96b766 100644
> --- a/kernel/utsname_sysctl.c
> +++ b/kernel/utsname_sysctl.c
> @@ -13,6 +13,7 @@
> #include <linux/uts.h>
> #include <linux/utsname.h>
> #include <linux/sysctl.h>
> +#include <linux/wait.h>
>
> static void *get_uts(ctl_table *table, int write)
> {
> @@ -51,12 +52,28 @@ static int proc_do_uts_string(ctl_table *table, int write,
> uts_table.data = get_uts(table, write);
> r = proc_dostring(&uts_table,write,buffer,lenp, ppos);
> put_uts(table, write, uts_table.data);
> +
> + if (write) {
> + atomic_inc(&table->poll->event);
> + wake_up_interruptible(&table->poll->wait);
Duplicating code again? Why are you not calling your generic notify?
> + }
> +
> return r;
> }
> #else
> #define proc_do_uts_string NULL
> #endif
>
> +static struct ctl_table_poll hostname_poll = {
> + .event = ATOMIC_INIT(0),
> + .wait = __WAIT_QUEUE_HEAD_INITIALIZER(hostname_poll.wait),
> +};
> +
> +static struct ctl_table_poll domainname_poll = {
> + .event = ATOMIC_INIT(0),
> + .wait = __WAIT_QUEUE_HEAD_INITIALIZER(domainname_poll.wait),
> +};
No generic initialize for this generic structure? Instead every user
must repeat the initialization?
> static struct ctl_table uts_kern_table[] = {
> {
> .procname = "ostype",
> @@ -85,6 +102,7 @@ static struct ctl_table uts_kern_table[] = {
> .maxlen = sizeof(init_uts_ns.name.nodename),
> .mode = 0644,
> .proc_handler = proc_do_uts_string,
> + .poll = &hostname_poll,
> },
> {
> .procname = "domainname",
> @@ -92,6 +110,7 @@ static struct ctl_table uts_kern_table[] = {
> .maxlen = sizeof(init_uts_ns.name.domainname),
> .mode = 0644,
> .proc_handler = proc_do_uts_string,
> + .poll = &domainname_poll,
> },
> {}
> };
> @@ -105,6 +124,23 @@ static struct ctl_table uts_root_table[] = {
> {}
> };
>
> +#ifdef CONFIG_PROC_SYSCTL
> +/*
> + * Notify userspace about a change in a certain entry of uts_kern_table,
> + * identified by the parameter proc.
> + */
> +void uts_proc_notify(enum uts_proc proc)
> +{
> + struct ctl_table *table = &uts_kern_table[proc];
> +
> + if (!table->poll)
> + return;
> +
> + atomic_inc(&table->poll->event);
> + wake_up_interruptible(&table->poll->wait);
Since you have a generic structure why don't you have a generic function
that calls the atomic_inc and wake_up_interruptible.
> +}
> +#endif
> +
> static int __init utsname_sysctl_init(void)
> {
> register_sysctl_table(uts_root_table);
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists