proc: make /proc/*/cmdline go through LSM
PATCH 1/5 proc: make /proc/*/cmdline go through LSM
/proc/*/cmdline is not different from /proc/*/environ as it accesses target task's memory (and can access the very same region of memory) but it doesn't go through ptrace_may_access() and thus doesn't go through LSM.
---
fs/proc/base.c | 67 ++++++++++++++++++++++++++++----------------------------- 1 file changed, 33 insertions(+), 34 deletions(-)
fs/proc/base.c +33 -34
@@ -208,11 +208,34 @@ static int proc_root_link(struct dentry *dentry, struct path *path)return result; } +static int __mem_open (struct inode *inode,struct file *file,unsigned int mode) +{ +struct mm_struct *mm =proc_mem_open (inode, mode); + +if (IS_ERR (mm)) +return PTR_ERR (mm); + + file->private_data = mm; +return 0 ; +} + +static int proc_pid_cmdline_open (struct inode *inode,struct file *file) +{ +return __mem_open (inode, file,PTRACE_MODE_READ ); +} + +static int mem_release (struct inode *inode,struct file *file) +{ +struct mm_struct *mm = file->private_data; +if (mm) +mmdrop (mm); +return 0 ; +} +static ssize_t proc_pid_cmdline_read (struct file *file,char __user *buf,size_t _count,loff_t *pos) { -struct task_struct *tsk; -struct mm_struct *mm; +struct mm_struct *mm = file->private_data;char *page;unsigned long count = _count;unsigned long arg_start, arg_end, env_start, env_end;
@@ -223,18 +246,11 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,BUG_ON (*pos <0 ); - tsk =get_proc_task (file_inode (file)); -if (!tsk) -return -ESRCH ; - mm =get_task_mm (tsk); -put_task_struct (tsk); -if (!mm) -return 0 ;/* Check if process spawned far enough to have cmdline. */ -if (!mm->env_end) { - rv =0 ; - goto out_mmput; - } +if (!mm || !mm->env_end) +return 0 ; +if (!mmget_not_zero (mm)) +return 0 ; page = (char *)__get_free_page (GFP_KERNEL );if (!page) {
@@ -376,8 +392,10 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, }static const struct file_operations proc_pid_cmdline_ops = { - .read = proc_pid_cmdline_read, - .llseek = generic_file_llseek, + .open = proc_pid_cmdline_open, + .read = proc_pid_cmdline_read, + .llseek = generic_file_llseek, + .release = mem_release, };#ifdef CONFIG_KALLSYMS
@@ -786,17 +804,6 @@ struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)return mm; } -static int __mem_open (struct inode *inode,struct file *file,unsigned int mode) -{ -struct mm_struct *mm =proc_mem_open (inode, mode); - -if (IS_ERR (mm)) -return PTR_ERR (mm); - - file->private_data = mm; -return 0 ; -} -static int mem_open (struct inode *inode,struct file *file) {int ret =__mem_open (inode, file,PTRACE_MODE_ATTACH );
@@ -890,14 +897,6 @@ loff_t mem_lseek(struct file *file, loff_t offset, int orig)return file->f_pos; } -static int mem_release (struct inode *inode,struct file *file) -{ -struct mm_struct *mm = file->private_data; -if (mm) -mmdrop (mm); -return 0 ; -} -static const struct file_operations proc_mem_operations = { .llseek = mem_lseek, .read = mem_read,
PATCH 2/5 proc: more "unsigned int" in /proc/*/cmdline
access_remote_vm() doesn't return negative errors, it returns number of bytes read/written (0 if error occurs). This allows to delete some comparisons which never trigger.
Reuse "nr_read" variable while I'm at it.
---
fs/proc/base.c | 29 ++++++++++++----------------- 1 file changed, 12 insertions(+), 17 deletions(-)
fs/proc/base.c +12 -17
@@ -280,9 +280,10 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,* Inherently racy -- command line shares address space * with code and data. */ - rv =access_remote_vm (mm, arg_end -1 , &c,1 ,0 ); -if (rv <=0 ) +if (access_remote_vm (mm, arg_end -1 , &c,1 ,0 ) !=1 ) { + rv =0 ; goto out_free_page; + } rv =0 ;
@@ -294,14 +295,11 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, p = arg_start + *pos; len = len1 - *pos;while (count >0 && len >0 ) { -unsigned int _count; -int nr_read; - - _count =min3 (count, len,PAGE_SIZE ); - nr_read =access_remote_vm (mm, p, page, _count,0 ); -if (nr_read <0 ) - rv = nr_read; -if (nr_read <=0 ) +unsigned int nr_read; + + nr_read =min3 (count, len,PAGE_SIZE ); + nr_read =access_remote_vm (mm, p, page, nr_read,0 ); +if (nr_read ==0 ) goto out_free_page;if (copy_to_user (buf, page, nr_read)) {
@@ -339,15 +337,12 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, p = cmdline[i].p + pos1; len = cmdline[i].len - pos1;while (count >0 && len >0 ) { -unsigned int _count, l; -int nr_read; +unsigned int nr_read, l;bool final; - _count =min3 (count, len,PAGE_SIZE ); - nr_read =access_remote_vm (mm, p, page, _count,0 ); -if (nr_read <0 ) - rv = nr_read; -if (nr_read <=0 ) + nr_read =min3 (count, len,PAGE_SIZE ); + nr_read =access_remote_vm (mm, p, page, nr_read,0 ); +if (nr_read ==0 ) goto out_free_page;/*
PATCH 3/5 proc: somewhat simpler code for /proc/*/cmdline
"final" variable is OK but we can get away with less lines.
---
fs/proc/base.c | 24 +++++++++--------------- 1 file changed, 9 insertions(+), 15 deletions(-)
fs/proc/base.c +9 -15
@@ -337,8 +337,7 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, p = cmdline[i].p + pos1; len = cmdline[i].len - pos1;while (count >0 && len >0 ) { -unsigned int nr_read, l; -bool final; +unsigned int nr_read, nr_write; nr_read =min3 (count, len,PAGE_SIZE ); nr_read =access_remote_vm (mm, p, page, nr_read,0 );
@@ -349,25 +348,20 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,* Command line can be shorter than whole ARGV * even if last "marker" byte says it is not. */ - final = false; - l =strnlen (page, nr_read); -if (l < nr_read) { - nr_read = l; - final = true; - } + nr_write =strnlen (page, nr_read); -if (copy_to_user (buf, page, nr_read)) { +if (copy_to_user (buf, page, nr_write)) { rv = -EFAULT ; goto out_free_page; } - p += nr_read; - len -= nr_read; - buf += nr_read; - count -= nr_read; - rv += nr_read; + p += nr_write; + len -= nr_write; + buf += nr_write; + count -= nr_write; + rv += nr_write; -if (final) +if (nr_write < nr_read) goto out_free_page; }
PATCH 4/5 proc: simpler iterations for /proc/*/cmdline
"rv" variable is used both as a counter of bytes transferred and an error value holder but it can be reduced solely to error values if original start of userspace buffer is stashed and used at the very end.
---
fs/proc/base.c | 36 +++++++++++++++++------------------- 1 file changed, 17 insertions(+), 19 deletions(-)
fs/proc/base.c +17 -19
@@ -241,8 +241,9 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,unsigned long arg_start, arg_end, env_start, env_end;unsigned long len1, len2, len;unsigned long p; +char __user *buf0 = buf;char c; -ssize_t rv; +int rv;BUG_ON (*pos <0 );
@@ -272,25 +273,20 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, len2 = env_end - env_start;/* Empty ARGV. */ -if (len1 ==0 ) { - rv =0 ; - goto out_free_page; - } +if (len1 ==0 ) + goto end; +/* * Inherently racy -- command line shares address space * with code and data. */ -if (access_remote_vm (mm, arg_end -1 , &c,1 ,0 ) !=1 ) { - rv =0 ; - goto out_free_page; - } - - rv =0 ; +if (access_remote_vm (mm, arg_end -1 , &c,1 ,0 ) !=1 ) + goto end;if (c =='\0' ) {/* Command line (set of strings) occupies whole ARGV. */ if (len1 <= *pos) - goto out_free_page; + goto end; p = arg_start + *pos; len = len1 - *pos;
@@ -300,7 +296,7 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, nr_read =min3 (count, len,PAGE_SIZE ); nr_read =access_remote_vm (mm, p, page, nr_read,0 );if (nr_read ==0 ) - goto out_free_page; + goto end;if (copy_to_user (buf, page, nr_read)) { rv = -EFAULT ;
@@ -311,7 +307,6 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, len -= nr_read; buf += nr_read; count -= nr_read; - rv += nr_read; } }else {/*
@@ -342,7 +337,7 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, nr_read =min3 (count, len,PAGE_SIZE ); nr_read =access_remote_vm (mm, p, page, nr_read,0 );if (nr_read ==0 ) - goto out_free_page; + goto end;/* * Command line can be shorter than whole ARGV
@@ -359,10 +354,9 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, len -= nr_write; buf += nr_write; count -= nr_write; - rv += nr_write;if (nr_write < nr_read) - goto out_free_page; + goto end; }/* Only first chunk can be read partially. */
@@ -371,12 +365,16 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, } } +end: +free_page ((unsigned long )page); +mmput (mm); + *pos += buf - buf0; +return buf - buf0; + out_free_page:free_page ((unsigned long )page); out_mmput:mmput (mm); -if (rv >0 ) - *pos += rv;return rv; }
PATCH 5/5 proc: deduplicate /proc/*/cmdline implementation
Code can be sonsolidated if a dummy region of 0 length is used in normal case of \0-separated command line:
1) [arg_start, arg_end) + [dummy len=0]
2) [arg_start, arg_end) + [env_start, env_end)
---
fs/proc/base.c | 53 ++++++++++++++++++++--------------------------------- 1 file changed, 20 insertions(+), 33 deletions(-)
fs/proc/base.c +20 -33
@@ -239,9 +239,12 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,char *page;unsigned long count = _count;unsigned long arg_start, arg_end, env_start, env_end; -unsigned long len1, len2, len; -unsigned long p; +unsigned long len1, len2;char __user *buf0 = buf; +struct { +unsigned long p; +unsigned long len; + } cmdline[2 ];char c;int rv;
@@ -283,43 +286,21 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,if (access_remote_vm (mm, arg_end -1 , &c,1 ,0 ) !=1 ) goto end; + cmdline[0 ].p = arg_start; + cmdline[0 ].len = len1;if (c =='\0' ) {/* Command line (set of strings) occupies whole ARGV. */ -if (len1 <= *pos) - goto end; - - p = arg_start + *pos; - len = len1 - *pos; -while (count >0 && len >0 ) { -unsigned int nr_read; - - nr_read =min3 (count, len,PAGE_SIZE ); - nr_read =access_remote_vm (mm, p, page, nr_read,0 ); -if (nr_read ==0 ) - goto end; - -if (copy_to_user (buf, page, nr_read)) { - rv = -EFAULT ; - goto out_free_page; - } - - p += nr_read; - len -= nr_read; - buf += nr_read; - count -= nr_read; - } + cmdline[1 ].len =0 ; }else {/* * Command line (1 string) occupies ARGV and * extends into ENVP. */ -struct { -unsigned long p; -unsigned long len; - } cmdline[2 ] = { - { .p = arg_start, .len = len1 }, - { .p = env_start, .len = len2 }, - }; + cmdline[1 ].p = env_start; + cmdline[1 ].len = len2; + } + + {loff_t pos1 = *pos;unsigned int i;
@@ -329,6 +310,9 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, i++; }while (i <2 ) { +unsigned long p; +unsigned long len; + p = cmdline[i].p + pos1; len = cmdline[i].len - pos1;while (count >0 && len >0 ) {
@@ -343,7 +327,10 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,* Command line can be shorter than whole ARGV * even if last "marker" byte says it is not. */ - nr_write =strnlen (page, nr_read); +if (c =='\0' ) + nr_write = nr_read; +else + nr_write =strnlen (page, nr_read);if (copy_to_user (buf, page, nr_write)) { rv = -EFAULT ;
Re: [PATCH 4/5] proc: simpler iterations for /proc/*/cmdline
In reply to Alexey DobriyanQuoted 20 lines · On Wed, 21 Feb 2018 22:30:09 +0300 Alexey Dobriyan <[email protected]> wrote:
"rv" variable is used both as a counter of bytes transferred and an error value holder but it can be reduced solely to error values if original start of userspace buffer is stashed and used at the very end.
...
fs/proc/base.c +4
-0@@ -371,12 +365,16 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, } } +end: +free_page ((unsigned long )page); +mmput (mm); + *pos += buf - buf0;
fs/proc/base.c static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, · line 379
Quoted 10 lines
fs/proc/base.c +2
-2+return buf - buf0; + out_free_page:free_page ((unsigned long )page); out_mmput:mmput (mm); -if (rv >0 ) - *pos += rv;return rv; }
(scratches head). Why not do this?
fs/proc/base.c renamed from fs/proc/base.c~proc-simpler-iterations-for-proc-cmdline-fix +1 -4
@@ -363,11 +363,8 @@ static ssize_t proc_pid_cmdline_read(str } end: -free_page ((unsigned long )page); -mmput (mm); *pos += buf - buf0; -return buf - buf0; - + rv = buf - buf0; out_free_page:free_page ((unsigned long )page); out_mmput:
(and maybe rv should be ssize_t)
Re: [PATCH 4/5] proc: simpler iterations for /proc/*/cmdline
Quoted 51 lines · Andrew Morton
Quoted 20 lines · On Wed, 21 Feb 2018 22:30:09 +0300 Alexey Dobriyan <[email protected]> wrote:
"rv" variable is used both as a counter of bytes transferred and an error value holder but it can be reduced solely to error values if original start of userspace buffer is stashed and used at the very end.
...
fs/proc/base.c +4
-0@@ -371,12 +365,16 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, } } +end: +free_page ((unsigned long )page); +mmput (mm); + *pos += buf - buf0;fs/proc/base.c static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, · line 379
Quoted 10 lines
fs/proc/base.c +2
-2+return buf - buf0; + out_free_page:free_page ((unsigned long )page); out_mmput:mmput (mm); -if (rv >0 ) - *pos += rv;return rv; }(scratches head). Why not do this?
fs/proc/base.c renamed from fs/proc/base.c~proc-simpler-iterations-for-proc-cmdline-fix +1
-4@@ -363,11 +363,8 @@ static ssize_t proc_pid_cmdline_read(str } end: -free_page ((unsigned long )page); -mmput (mm); *pos += buf - buf0; -return buf - buf0; - + rv = buf - buf0; out_free_page:free_page ((unsigned long )page); out_mmput:
Quoted 1 line
(and maybe rv should be
ssize_t)
It should as cmdline area can be set to arbitrary values.
I hate ssize_t and it's bloat.
$ ./scripts/bloat-o-meter ../vmlinux-000 ../obj/vmlinux add/remove: 0/0 grow/shrink: 1/0 up/down: 11/0 (11) Function old new delta proc_pid_cmdline_read 867 878 +11
fs/proc/base.c static int proc_root_link(struct dentry *dentry, struct path *path) · line 216
Quoted 10 lines · Alexey Dobriyan
/proc/*/cmdlineis not different from/proc/*/environas it accesses target task's memory (and can access the very same region of memory) but it doesn't go throughptrace_may_access()and thus doesn't go through LSM.fs/proc/base.c +6
-0+static int __mem_open (struct inode *inode,struct file *file,unsigned int mode) +{ +struct mm_struct *mm =proc_mem_open (inode, mode); + +if (IS_ERR (mm)) +return PTR_ERR (mm);
So, is it possible to have it NULL?..
fs/proc/base.c static int proc_root_link(struct dentry *dentry, struct path *path) · line 230
Quoted 10 lines
fs/proc/base.c +8
-0+ + file->private_data = mm; +return 0 ; +} +static int mem_release (struct inode *inode,struct file *file) +{ +struct mm_struct *mm = file->private_data; +if (mm)
...or I don't get this check.
Quoted 3 lines
fs/proc/base.c +3
-0+mmdrop (mm); +return 0 ; +}
Signature · 2 lines
With Best Regards, Andy Shevchenko
Quoted 5 lines · Andy Shevchenko
Quoted 3 lines · Alexey Dobriyan
/proc/*/cmdlineis not different from/proc/*/environas it accesses target task's memory (and can access the very same region of memory) but it doesn't go throughptrace_may_access()and thus doesn't go through LSM.
Quoted 8 lines
fs/proc/base.c static int proc_root_link(struct dentry *dentry, struct path *path) · line 216
Quoted 6 lines
fs/proc/base.c +6
-0+static int __mem_open (struct inode *inode,struct file *file,unsigned int mode) +{ +struct mm_struct *mm =proc_mem_open (inode, mode); + +if (IS_ERR (mm)) +return PTR_ERR (mm);So, is it possible to have it NULL?..
I haven't looked closely, but if kernel thread is accesses then yes.
Regardless, patch only moves function so that code compiles, untangling this little mess is separate adventure.
Quoted 11 lines
fs/proc/base.c static int proc_root_link(struct dentry *dentry, struct path *path) · line 230
Quoted 5 lines
fs/proc/base.c +4
-0+static int mem_release (struct inode *inode,struct file *file) +{ +struct mm_struct *mm = file->private_data; +if (mm)...or I don't get this check.
Quoted 3 lines
fs/proc/base.c +3
-0+mmdrop (mm); +return 0 ; +}
It should trigger if kernel thread is accessed.
Quoted 4 lines · On Wed, 21 Feb 2018 22:23:39 +0300 Alexey Dobriyan <[email protected]> wrote:
/proc/*/cmdlineis not different from/proc/*/environas it accesses target task's memory (and can access the very same region of memory) but it doesn't go throughptrace_may_access()and thus doesn't go through LSM.
I'd really like to see more thoughtful changelogging, please.
Why are we doing this? What is the advantage?
Doesn't this mean that code which could previously read /proc/pid/cmdline may no longer be able to do so? Can't this break userspace? Discuss. Lots!
PATCH v2 1/5 proc: make /proc/*/cmdline go through LSM
/proc/*/mem and /proc/*/environ use common code to check if open is allowed. Essentially they all access target task memory.
/proc/*/cmdline should be subject to the same checks as it does what /proc/*/environ does.
->security_ptrace_access_check LSM hook will be called allowing LSM to interfere.
---
fs/proc/base.c | 67 ++++++++++++++++++++++++++++----------------------------- 1 file changed, 33 insertions(+), 34 deletions(-)
fs/proc/base.c +33 -34
@@ -208,11 +208,34 @@ static int proc_root_link(struct dentry *dentry, struct path *path)return result; } +static int __mem_open (struct inode *inode,struct file *file,unsigned int mode) +{ +struct mm_struct *mm =proc_mem_open (inode, mode); + +if (IS_ERR (mm)) +return PTR_ERR (mm); + + file->private_data = mm; +return 0 ; +} + +static int proc_pid_cmdline_open (struct inode *inode,struct file *file) +{ +return __mem_open (inode, file,PTRACE_MODE_READ ); +} + +static int mem_release (struct inode *inode,struct file *file) +{ +struct mm_struct *mm = file->private_data; +if (mm) +mmdrop (mm); +return 0 ; +} +static ssize_t proc_pid_cmdline_read (struct file *file,char __user *buf,size_t _count,loff_t *pos) { -struct task_struct *tsk; -struct mm_struct *mm; +struct mm_struct *mm = file->private_data;char *page;unsigned long count = _count;unsigned long arg_start, arg_end, env_start, env_end;
@@ -223,18 +246,11 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,BUG_ON (*pos <0 ); - tsk =get_proc_task (file_inode (file)); -if (!tsk) -return -ESRCH ; - mm =get_task_mm (tsk); -put_task_struct (tsk); -if (!mm) -return 0 ;/* Check if process spawned far enough to have cmdline. */ -if (!mm->env_end) { - rv =0 ; - goto out_mmput; - } +if (!mm || !mm->env_end) +return 0 ; +if (!mmget_not_zero (mm)) +return 0 ; page = (char *)__get_free_page (GFP_KERNEL );if (!page) {
@@ -376,8 +392,10 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, }static const struct file_operations proc_pid_cmdline_ops = { - .read = proc_pid_cmdline_read, - .llseek = generic_file_llseek, + .open = proc_pid_cmdline_open, + .read = proc_pid_cmdline_read, + .llseek = generic_file_llseek, + .release = mem_release, };#ifdef CONFIG_KALLSYMS
@@ -786,17 +804,6 @@ struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)return mm; } -static int __mem_open (struct inode *inode,struct file *file,unsigned int mode) -{ -struct mm_struct *mm =proc_mem_open (inode, mode); - -if (IS_ERR (mm)) -return PTR_ERR (mm); - - file->private_data = mm; -return 0 ; -} -static int mem_open (struct inode *inode,struct file *file) {int ret =__mem_open (inode, file,PTRACE_MODE_ATTACH );
@@ -890,14 +897,6 @@ loff_t mem_lseek(struct file *file, loff_t offset, int orig)return file->f_pos; } -static int mem_release (struct inode *inode,struct file *file) -{ -struct mm_struct *mm = file->private_data; -if (mm) -mmdrop (mm); -return 0 ; -} -static const struct file_operations proc_mem_operations = { .llseek = mem_lseek, .read = mem_read,
Quoted 3 lines · On Wed, 21 Feb 2018 22:23:39 +0300 Alexey Dobriyan <[email protected]> wrote:
/proc/*/cmdlineis not different from/proc/*/environas it accesses target task's memory (and can access the very same region of memory) but it doesn't go throughptrace_may_access()and thus doesn't go through LSM.
This change can cause existing code to break, no?
I'd like to see, in the changelog, a full explanation of why this won't break any existing setup?
Quoted 10 lines · Andrew Morton
Quoted 3 lines · On Wed, 21 Feb 2018 22:23:39 +0300 Alexey Dobriyan <[email protected]> wrote:
/proc/*/cmdlineis not different from/proc/*/environas it accesses target task's memory (and can access the very same region of memory) but it doesn't go throughptrace_may_access()and thus doesn't go through LSM.This change can cause existing code to break, no?
I'd like to see, in the changelog, a full explanation of why this won't break any existing setup?
It can. In fact, I'm not sure about this patch anymore. Original /proc/*/cmdline hook did get_cmdline() which is basically GUP. It is just nobody said anything when /proc/*/cmdline got rewritten and security folks aren't bitching about it:
$ cat /proc/1/cmdline init [3]
I'll resend the rest of cmdline changes if they gets broken.