From: "David S. Miller" The issue is that proc_dointvec_jiffies does not normalize to/from USER_HZ, and there are a bunch of networking bits that need this. Begin forwarded message: Date: Thu, 18 Dec 2003 15:42:27 +0900 (JST) From: YOSHIFUJI Hideaki / _$B5HF#1QL@ To: davem@redhat.com Cc: netdev@oss.sgi.com, yoshfuji@linux-ipv6.org Subject: Re: [PATCH] NET: Normalize jiffies reported to userspace, in neighbor management code Hello. In article <20031111160200.72cee93a.davem@redhat.com> (at Tue, 11 Nov 2003 16:02:00 -0800), "David S. Miller" says: > On Tue, 11 Nov 2003 16:31:28 -0600 (CST) > YOSHIFUJI Hideaki / _$B5HF#1QL@ wrote: > > > In article <20031110230233.254061da.davem@redhat.com> (at Mon, 10 Nov 2003 23:02:33 -0800), "David S. Miller" says: > > > > > Another idea is to change do_proc_dointvec() to take a conversion function > > > pointer instead of this "conv" thing. Maybe even proc_dointvec_minmax() > > > could be implemented in terms of do_proc_dointvec() with such a scheme. > > > > This is essentially identical to what I thought. > > Okay, how about this? > > This looks very nice, thank you Yoshfuji. > > I'm asking Linus what we should do here. It's a big change to > make to fix this bug, and if he thinks so too we'll just use > your original patch which is a safer version of the fix for 2.6.0 > purposes. If that happens, we'll integrate this nicer version > for 2.6.1 or something like this. 2.6.0 is out finally; it would be the good time for resending the patch. Thanks you in advance. ===== include/linux/sysctl.h 1.53 vs edited ===== --- include/linux/sysctl.h | 2 kernel/sysctl.c | 327 ++++++++++++++++++++++++++++++------------------- net/core/neighbour.c | 9 - 3 files changed, 212 insertions(+), 126 deletions(-) diff -puN include/linux/sysctl.h~net-jiffy-normalisation-fix include/linux/sysctl.h --- 25/include/linux/sysctl.h~net-jiffy-normalisation-fix 2004-01-19 15:37:55.000000000 -0800 +++ 25-akpm/include/linux/sysctl.h 2004-01-19 15:37:55.000000000 -0800 @@ -737,6 +737,8 @@ extern int proc_dointvec_minmax(ctl_tabl void __user *, size_t *); extern int proc_dointvec_jiffies(ctl_table *, int, struct file *, void __user *, size_t *); +extern int proc_dointvec_userhz_jiffies(ctl_table *, int, struct file *, + void __user *, size_t *); extern int proc_doulongvec_minmax(ctl_table *, int, struct file *, void __user *, size_t *); extern int proc_doulongvec_ms_jiffies_minmax(ctl_table *table, int, diff -puN kernel/sysctl.c~net-jiffy-normalisation-fix kernel/sysctl.c --- 25/kernel/sysctl.c~net-jiffy-normalisation-fix 2004-01-19 15:37:55.000000000 -0800 +++ 25-akpm/kernel/sysctl.c 2004-01-19 15:37:55.000000000 -0800 @@ -37,6 +37,7 @@ #include #include #include +#include #include #ifdef CONFIG_ROOT_NFS @@ -1070,8 +1071,8 @@ int do_sysctl_strategy (ctl_table *table * cover common cases - * * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(), - * proc_dointvec_minmax(), proc_doulongvec_ms_jiffies_minmax(), - * proc_doulongvec_minmax() + * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(), + * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax() * * It is the handler's job to read the input buffer from user memory * and process it. The handler should return 0 on success. @@ -1342,19 +1343,36 @@ static int proc_doutsstring(ctl_table *t return r; } -#define OP_SET 0 -#define OP_AND 1 -#define OP_OR 2 -#define OP_MAX 3 -#define OP_MIN 4 +static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp, + int *valp, + int write, void *data) +{ + if (write) { + *valp = *negp ? -*lvalp : *lvalp; + } else { + int val = *valp; + if (val < 0) { + *negp = -1; + *lvalp = (unsigned long)-val; + } else { + *negp = 0; + *lvalp = (unsigned long)val; + } + } + return 0; +} static int do_proc_dointvec(ctl_table *table, int write, struct file *filp, - void __user *buffer, size_t *lenp, int conv, int op) + void __user *buffer, size_t *lenp, + int (*conv)(int *negp, unsigned long *lvalp, int *valp, + int write, void *data), + void *data) { +#define TMPBUFLEN 20 int *i, vleft, first=1, neg, val; + unsigned long lval; size_t left, len; - #define TMPBUFLEN 20 char buf[TMPBUFLEN], *p; if (!table->data || !table->maxlen || !*lenp || @@ -1364,9 +1382,12 @@ static int do_proc_dointvec(ctl_table *t } i = (int *) table->data; - vleft = table->maxlen / sizeof(int); + vleft = table->maxlen / sizeof(*i); left = *lenp; - + + if (!conv) + conv = do_proc_dointvec_conv; + for (; left && vleft--; i++, first=0) { if (write) { while (left) { @@ -1382,8 +1403,8 @@ static int do_proc_dointvec(ctl_table *t break; neg = 0; len = left; - if (len > TMPBUFLEN-1) - len = TMPBUFLEN-1; + if (len > sizeof(buf) - 1) + len = sizeof(buf) - 1; if(copy_from_user(buf, buffer, len)) return -EFAULT; buf[len] = 0; @@ -1394,7 +1415,9 @@ static int do_proc_dointvec(ctl_table *t } if (*p < '0' || *p > '9') break; - val = simple_strtoul(p, &p, 0) * conv; + + lval = simple_strtoul(p, &p, 0); + len = p-buf; if ((len < left) && *p && !isspace(*p)) break; @@ -1402,22 +1425,18 @@ static int do_proc_dointvec(ctl_table *t val = -val; buffer += len; left -= len; - switch(op) { - case OP_SET: *i = val; break; - case OP_AND: *i &= val; break; - case OP_OR: *i |= val; break; - case OP_MAX: if(*i < val) - *i = val; - break; - case OP_MIN: if(*i > val) - *i = val; - break; - } + + if (conv(&neg, &lval, i, 1, data)) + break; } else { p = buf; if (!first) *p++ = '\t'; - sprintf(p, "%d", (*i) / conv); + + if (conv(&neg, &lval, i, 0, data)) + break; + + sprintf(p, "%s%lu", neg ? "-" : "", lval); len = strlen(buf); if (len > left) len = left; @@ -1449,6 +1468,7 @@ static int do_proc_dointvec(ctl_table *t *lenp -= left; filp->f_pos += *lenp; return 0; +#undef TMPBUFLEN } /** @@ -1467,7 +1487,45 @@ static int do_proc_dointvec(ctl_table *t int proc_dointvec(ctl_table *table, int write, struct file *filp, void __user *buffer, size_t *lenp) { - return do_proc_dointvec(table,write,filp,buffer,lenp,1,OP_SET); + return do_proc_dointvec(table,write,filp,buffer,lenp, + NULL,NULL); +} + +#define OP_SET 0 +#define OP_AND 1 +#define OP_OR 2 +#define OP_MAX 3 +#define OP_MIN 4 + +static int do_proc_dointvec_bset_conv(int *negp, unsigned long *lvalp, + int *valp, + int write, void *data) +{ + int op = *(int *)data; + if (write) { + int val = *negp ? -*lvalp : *lvalp; + switch(op) { + case OP_SET: *valp = val; break; + case OP_AND: *valp &= val; break; + case OP_OR: *valp |= val; break; + case OP_MAX: if(*valp < val) + *valp = val; + break; + case OP_MIN: if(*valp > val) + *valp = val; + break; + } + } else { + int val = *valp; + if (val < 0) { + *negp = -1; + *lvalp = (unsigned long)-val; + } else { + *negp = 0; + *lvalp = (unsigned long)val; + } + } + return 0; } /* @@ -1477,11 +1535,44 @@ int proc_dointvec(ctl_table *table, int int proc_dointvec_bset(ctl_table *table, int write, struct file *filp, void __user *buffer, size_t *lenp) { + int op; + if (!capable(CAP_SYS_MODULE)) { return -EPERM; } - return do_proc_dointvec(table,write,filp,buffer,lenp,1, - (current->pid == 1) ? OP_SET : OP_AND); + + op = (current->pid == 1) ? OP_SET : OP_AND; + return do_proc_dointvec(table,write,filp,buffer,lenp, + do_proc_dointvec_bset_conv,&op); +} + +struct do_proc_dointvec_minmax_conv_param { + int *min; + int *max; +}; + +static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp, + int *valp, + int write, void *data) +{ + struct do_proc_dointvec_minmax_conv_param *param = data; + if (write) { + int val = *negp ? -*lvalp : *lvalp; + if ((param->min && *param->min > val) || + (param->max && *param->max < val)) + return -EINVAL; + *valp = val; + } else { + int val = *valp; + if (val < 0) { + *negp = -1; + *lvalp = (unsigned long)-val; + } else { + *negp = 0; + *lvalp = (unsigned long)val; + } + } + return 0; } /** @@ -1503,98 +1594,12 @@ int proc_dointvec_bset(ctl_table *table, int proc_dointvec_minmax(ctl_table *table, int write, struct file *filp, void __user *buffer, size_t *lenp) { - int *i, *min, *max, vleft, first=1, neg, val; - size_t len, left; - #define TMPBUFLEN 20 - char buf[TMPBUFLEN], *p; - - if (!table->data || !table->maxlen || !*lenp || - (filp->f_pos && !write)) { - *lenp = 0; - return 0; - } - - i = (int *) table->data; - min = (int *) table->extra1; - max = (int *) table->extra2; - vleft = table->maxlen / sizeof(int); - left = *lenp; - - for (; left && vleft--; i++, min++, max++, first=0) { - if (write) { - while (left) { - char c; - if(get_user(c, (char *) buffer)) - return -EFAULT; - if (!isspace(c)) - break; - left--; - buffer++; - } - if (!left) - break; - neg = 0; - len = left; - if (len > TMPBUFLEN-1) - len = TMPBUFLEN-1; - if(copy_from_user(buf, buffer, len)) - return -EFAULT; - buf[len] = 0; - p = buf; - if (*p == '-' && left > 1) { - neg = 1; - left--, p++; - } - if (*p < '0' || *p > '9') - break; - val = simple_strtoul(p, &p, 0); - len = p-buf; - if ((len < left) && *p && !isspace(*p)) - break; - if (neg) - val = -val; - buffer += len; - left -= len; - - if ((min && val < *min) || (max && val > *max)) - continue; - *i = val; - } else { - p = buf; - if (!first) - *p++ = '\t'; - sprintf(p, "%d", *i); - len = strlen(buf); - if (len > left) - len = left; - if(copy_to_user(buffer, buf, len)) - return -EFAULT; - left -= len; - buffer += len; - } - } - - if (!write && !first && left) { - if(put_user('\n', (char *) buffer)) - return -EFAULT; - left--, buffer++; - } - if (write) { - p = (char *) buffer; - while (left) { - char c; - if(get_user(c, p++)) - return -EFAULT; - if (!isspace(c)) - break; - left--; - } - } - if (write && first) - return -EINVAL; - *lenp -= left; - filp->f_pos += *lenp; - return 0; + struct do_proc_dointvec_minmax_conv_param param = { + .min = (int *) table->extra1, + .max = (int *) table->extra2, + }; + return do_proc_dointvec(table, write, filp, buffer, lenp, + do_proc_dointvec_minmax_conv, ¶m); } static int do_proc_doulongvec_minmax(ctl_table *table, int write, @@ -1749,6 +1754,48 @@ int proc_doulongvec_ms_jiffies_minmax(ct } +static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp, + int *valp, + int write, void *data) +{ + if (write) { + *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ); + } else { + int val = *valp; + unsigned long lval; + if (val < 0) { + *negp = -1; + lval = (unsigned long)-val; + } else { + *negp = 0; + lval = (unsigned long)val; + } + *lvalp = lval / HZ; + } + return 0; +} + +static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp, + int *valp, + int write, void *data) +{ + if (write) { + *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp); + } else { + int val = *valp; + unsigned long lval; + if (val < 0) { + *negp = -1; + lval = (unsigned long)-val; + } else { + *negp = 0; + lval = (unsigned long)val; + } + *lvalp = jiffies_to_clock_t(lval); + } + return 0; +} + /** * proc_dointvec_jiffies - read a vector of integers as seconds * @table: the sysctl table @@ -1767,7 +1814,30 @@ int proc_doulongvec_ms_jiffies_minmax(ct int proc_dointvec_jiffies(ctl_table *table, int write, struct file *filp, void __user *buffer, size_t *lenp) { - return do_proc_dointvec(table,write,filp,buffer,lenp,HZ,OP_SET); + return do_proc_dointvec(table,write,filp,buffer,lenp, + do_proc_dointvec_jiffies_conv,NULL); +} + +/** + * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds + * @table: the sysctl table + * @write: %TRUE if this is a write to the sysctl file + * @filp: the file structure + * @buffer: the user buffer + * @lenp: the size of the user buffer + * + * Reads/writes up to table->maxlen/sizeof(unsigned int) integer + * values from/to the user buffer, treated as an ASCII string. + * The values read are assumed to be in 1/USER_HZ seconds, and + * are converted into jiffies. + * + * Returns 0 on success. + */ +int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp, + void __user *buffer, size_t *lenp) +{ + return do_proc_dointvec(table,write,filp,buffer,lenp, + do_proc_dointvec_userhz_jiffies_conv,NULL); } #else /* CONFIG_PROC_FS */ @@ -1808,6 +1878,12 @@ int proc_dointvec_jiffies(ctl_table *tab return -ENOSYS; } +int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp, + void *buffer, size_t *lenp) +{ + return -ENOSYS; +} + int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp, void *buffer, size_t *lenp) { @@ -1995,6 +2071,12 @@ int proc_dointvec_jiffies(ctl_table *tab return -ENOSYS; } +int proc_dointvec_userhz_jiffies(ctl_table *table, int write, struct file *filp, + void *buffer, size_t *lenp) +{ + return -ENOSYS; +} + int proc_doulongvec_minmax(ctl_table *table, int write, struct file *filp, void __user *buffer, size_t *lenp) { @@ -2027,6 +2109,7 @@ void unregister_sysctl_table(struct ctl_ EXPORT_SYMBOL(proc_dointvec); EXPORT_SYMBOL(proc_dointvec_jiffies); EXPORT_SYMBOL(proc_dointvec_minmax); +EXPORT_SYMBOL(proc_dointvec_userhz_jiffies); EXPORT_SYMBOL(proc_dostring); EXPORT_SYMBOL(proc_doulongvec_minmax); EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax); diff -puN net/core/neighbour.c~net-jiffy-normalisation-fix net/core/neighbour.c --- 25/net/core/neighbour.c~net-jiffy-normalisation-fix 2004-01-19 15:37:55.000000000 -0800 +++ 25-akpm/net/core/neighbour.c 2004-01-19 15:37:55.000000000 -0800 @@ -24,6 +24,7 @@ #ifdef CONFIG_SYSCTL #include #endif +#include #include #include #include @@ -1510,7 +1511,7 @@ struct neigh_sysctl_table { .procname = "retrans_time", .maxlen = sizeof(int), .mode = 0644, - .proc_handler = &proc_dointvec, + .proc_handler = &proc_dointvec_userhz_jiffies, }, { .ctl_name = NET_NEIGH_REACHABLE_TIME, @@ -1555,21 +1556,21 @@ struct neigh_sysctl_table { .procname = "anycast_delay", .maxlen = sizeof(int), .mode = 0644, - .proc_handler = &proc_dointvec, + .proc_handler = &proc_dointvec_userhz_jiffies, }, { .ctl_name = NET_NEIGH_PROXY_DELAY, .procname = "proxy_delay", .maxlen = sizeof(int), .mode = 0644, - .proc_handler = &proc_dointvec, + .proc_handler = &proc_dointvec_userhz_jiffies, }, { .ctl_name = NET_NEIGH_LOCKTIME, .procname = "locktime", .maxlen = sizeof(int), .mode = 0644, - .proc_handler = &proc_dointvec, + .proc_handler = &proc_dointvec_userhz_jiffies, }, { .ctl_name = NET_NEIGH_GC_INTERVAL, _