Re: [RFC PATCH] vsprintf: Add %pv extension replacement for print_vma_addr
On Mon, 2020-08-17 at 14:44 +0300, Andy Shevchenko wrote: > On Fri, Aug 14, 2020 at 10:53:03AM -0700, Joe Perches wrote: > > There is a print_vma_addr function used several places > > that requires KERN_CONT use. > > > > Add a %pv mechanism to avoid the need for KERN_CONT. > > I like the idea, but I would accent the selling point to make code > (in the user call sites) nicer. [] > > +#else > > + buf = string_nocheck(buf, end, "CONFIG_MMU=n", default_str_spec); > > +#endif > > Can we avoid this spammy message? Really it's quite long and fills valuable > space in kernel buffer. I would rather print the hashed pointer as it's done > for the rest of %pXXX. Don't see why not. I believe it'd just use buf = ptr_to_id(buf, end, ip, spec); instead
Re: [RFC PATCH] vsprintf: Add %pv extension replacement for print_vma_addr
On Fri, Aug 14, 2020 at 10:53:03AM -0700, Joe Perches wrote:
> There is a print_vma_addr function used several places
> that requires KERN_CONT use.
>
> Add a %pv mechanism to avoid the need for KERN_CONT.
I like the idea, but I would accent the selling point to make code
(in the user call sites) nicer.
> An example conversion is arch/x86/kernel/signal.c
>
> from:
> if (show_unhandled_signals && printk_ratelimit()) {
> printk("%s"
> "%s[%d] bad frame in %s frame:%p ip:%lx sp:%lx orax:%lx",
> task_pid_nr(current) > 1 ? KERN_INFO : KERN_EMERG,
> me->comm, me->pid, where, frame,
> regs->ip, regs->sp, regs->orig_ax);
> print_vma_addr(KERN_CONT " in ", regs->ip);
> pr_cont("\n");
> to:
> printk("%s"
> "%s[%d] bad frame in %s frame:%p ip:%lx sp:%lx orax:%lx
> in %pv\n",
> task_pid_nr(current) > 1 ? KERN_INFO : KERN_EMERG,
> me->comm, me->pid, where, frame,
> regs->ip, regs->sp, regs->orig_ax, (void *)regs->ip);
...
> +#else
> + buf = string_nocheck(buf, end, "CONFIG_MMU=n", default_str_spec);
> +#endif
Can we avoid this spammy message? Really it's quite long and fills valuable
space in kernel buffer. I would rather print the hashed pointer as it's done
for the rest of %pXXX.
--
With Best Regards,
Andy Shevchenko
Re: [RFC PATCH] vsprintf: Add %pv extension replacement for print_vma_addr
On Sat, 2020-08-15 at 12:52 +0900, Sergey Senozhatsky wrote:
> Cc-ing John
>
> On (20/08/14 10:53), Joe Perches wrote:
> [..]
>
> In general, the idea looks nice.
>
> > +static noinline_for_stack
> > +char *vma_addr(char *buf, char *end, void *ip,
> > + struct printf_spec spec, const char *fmt)
> > +{
> > +#ifdef CONFIG_MMU
> > + struct mm_struct *mm = current->mm;
> > + struct vm_area_struct *vma;
> > +
> > + /*
> > +* we might be running from an atomic context so we cannot sleep
> > +*/
> > + if (!mmap_read_trylock(mm))
> > + return buf;
> > +
> > + vma = find_vma(mm, (unsigned long)ip);
> > + if (vma && vma->vm_file) {
> > + char *p;
> > + struct file *f = vma->vm_file;
> > + char *page = (char *)__get_free_page(GFP_NOWAIT);
>
> Hmm, this is huge. For the time being this is going to introduce lock
> inversion chains:
>
> PRINTK -> printk_locks -> MM -> mm_locks
>
> vs
> MM -> mm_locks -> PRINTK -> printk_locks
>
> Another thing to mention is
>
> PRINTK -> printk_locks -> MM (WANR_ON/etc) -> PRINTK
>
> we are in printk_safe, currently, but that's going to change.
>
> We might not be ready to take this as of now, but things can change
> once we drop printk_locks.
>
> > + if (page) {
> > + p = file_path(f, page, PAGE_SIZE);
> > + if (IS_ERR(p))
> > + p = "?";
> > + buf = string(buf, end, kbasename(p), default_str_spec);
> > + buf = string_nocheck(buf, end, "[", default_str_spec);
> > + buf = pointer_string(buf, end,
> > +(void *)vma->vm_start,
> > +default_hex_spec);
> > + buf = string_nocheck(buf, end, "+", default_str_spec);
> > + buf = pointer_string(buf, end,
> > +(void *)(vma->vm_end -
> > vma->vm_start),
> > +default_hex_spec);
> > + buf = string_nocheck(buf, end, "]", default_str_spec);
> > + free_page((unsigned long)page);
> > + }
> > + }
> > + mmap_read_unlock(mm);
> > +#else
> > + buf = string_nocheck(buf, end, "CONFIG_MMU=n", default_str_spec);
>
> Hmm, we don't usually do that CONFIG_FOO=n thing, I think we just need
> to skip %pv and print nothing.
I don't.
Right now, include/linux/mm.h has
#ifdef CONFIG_MMU
void print_vma_addr(char *prefix, unsigned long rip);
#else
static inline void print_vma_addr(char *prefix, unsigned long rip)
{
}
#endif
and the 'v' can't be #ifdef'd in vsprintf/pointer()
otherwise %pv would fallthrough to
return ptr_to_id(buf, end, ptr, spec);
Re: [RFC PATCH] vsprintf: Add %pv extension replacement for print_vma_addr
Cc-ing John
On (20/08/14 10:53), Joe Perches wrote:
[..]
In general, the idea looks nice.
> +static noinline_for_stack
> +char *vma_addr(char *buf, char *end, void *ip,
> +struct printf_spec spec, const char *fmt)
> +{
> +#ifdef CONFIG_MMU
> + struct mm_struct *mm = current->mm;
> + struct vm_area_struct *vma;
> +
> + /*
> + * we might be running from an atomic context so we cannot sleep
> + */
> + if (!mmap_read_trylock(mm))
> + return buf;
> +
> + vma = find_vma(mm, (unsigned long)ip);
> + if (vma && vma->vm_file) {
> + char *p;
> + struct file *f = vma->vm_file;
> + char *page = (char *)__get_free_page(GFP_NOWAIT);
Hmm, this is huge. For the time being this is going to introduce lock
inversion chains:
PRINTK -> printk_locks -> MM -> mm_locks
vs
MM -> mm_locks -> PRINTK -> printk_locks
Another thing to mention is
PRINTK -> printk_locks -> MM (WANR_ON/etc) -> PRINTK
we are in printk_safe, currently, but that's going to change.
We might not be ready to take this as of now, but things can change
once we drop printk_locks.
> + if (page) {
> + p = file_path(f, page, PAGE_SIZE);
> + if (IS_ERR(p))
> + p = "?";
> + buf = string(buf, end, kbasename(p), default_str_spec);
> + buf = string_nocheck(buf, end, "[", default_str_spec);
> + buf = pointer_string(buf, end,
> + (void *)vma->vm_start,
> + default_hex_spec);
> + buf = string_nocheck(buf, end, "+", default_str_spec);
> + buf = pointer_string(buf, end,
> + (void *)(vma->vm_end -
> vma->vm_start),
> + default_hex_spec);
> + buf = string_nocheck(buf, end, "]", default_str_spec);
> + free_page((unsigned long)page);
> + }
> + }
> + mmap_read_unlock(mm);
> +#else
> + buf = string_nocheck(buf, end, "CONFIG_MMU=n", default_str_spec);
Hmm, we don't usually do that CONFIG_FOO=n thing, I think we just need
to skip %pv and print nothing. Otherwise on !CONFIG_MMU systems the logbuf
may contain a number of CONFIG_MMU=n messages, which are hardly useful.
> +#endif
> + return buf;
> +}
> +
> /*
> * Show a '%p' thing. A kernel extension is that the '%p' is followed
> * by an extra set of alphanumeric characters that are extended format
> @@ -2254,6 +2304,8 @@ char *pointer(const char *fmt, char *buf, char *end,
> void *ptr,
> return uuid_string(buf, end, ptr, spec, fmt);
> case 'V':
> return va_format(buf, end, ptr, spec, fmt);
+ #ifdef CONFIG_MMU
> + case 'v':
> + return vma_addr(buf, end, ptr, spec, fmt);
+ #endif
> case 'K':
> return restricted_pointer(buf, end, ptr, spec);
> case 'N':
-ss
Re: [RFC PATCH] vsprintf: Add %pv extension replacement for print_vma_addr
On Fri, 2020-08-14 at 14:38 -0400, Steven Rostedt wrote:
> On Fri, 14 Aug 2020 10:53:03 -0700
> Joe Perches wrote:
> I'm fine with all his, but I feel more comfortable if this patch
> created a single copy of the code. Perhaps add:
[]
> diff --git a/mm/memory.c b/mm/memory.c
[]
> @@ -4771,32 +4771,7 @@ EXPORT_SYMBOL_GPL(access_process_vm);
> */
> void print_vma_addr(char *prefix, unsigned long ip)
> {
> - struct mm_struct *mm = current->mm;
> - struct vm_area_struct *vma;
> -
> - /*
> - * we might be running from an atomic context so we cannot sleep
> - */
> - if (!mmap_read_trylock(mm))
> - return;
> -
> - vma = find_vma(mm, ip);
> - if (vma && vma->vm_file) {
> - struct file *f = vma->vm_file;
> - char *buf = (char *)__get_free_page(GFP_NOWAIT);
> - if (buf) {
> - char *p;
> -
> - p = file_path(f, buf, PAGE_SIZE);
> - if (IS_ERR(p))
> - p = "?";
> - printk("%s%s[%lx+%lx]", prefix, kbasename(p),
> - vma->vm_start,
> - vma->vm_end - vma->vm_start);
> - free_page((unsigned long)buf);
> - }
> - }
> - mmap_read_unlock(mm);
> + printk("%s%pv", prefix, ip);
> }
I'd just convert all of them and remove this function instead.
Something like (uncompiled/untested):
---
arch/arc/kernel/troubleshoot.c | 2 +-
arch/arm64/kernel/traps.c | 16 +++--
arch/csky/kernel/traps.c | 11 -
arch/mips/mm/fault.c | 14 +--
arch/parisc/mm/fault.c | 15 +---
arch/powerpc/kernel/traps.c| 8 ++-
arch/riscv/kernel/traps.c | 11 -
arch/s390/mm/fault.c | 7 +++---
arch/sparc/mm/fault_32.c | 8 ++-
arch/sparc/mm/fault_64.c | 8 ++-
arch/um/kernel/trap.c | 13 ---
arch/x86/kernel/signal.c | 6 ++---
arch/x86/kernel/traps.c| 6 ++---
arch/x86/mm/fault.c| 53 +++---
include/linux/mm.h | 7 --
mm/memory.c| 33 --
16 files changed, 74 insertions(+), 144 deletions(-)
diff --git a/arch/arc/kernel/troubleshoot.c b/arch/arc/kernel/troubleshoot.c
index 28e8bf04b253..26ecd59f0057 100644
--- a/arch/arc/kernel/troubleshoot.c
+++ b/arch/arc/kernel/troubleshoot.c
@@ -86,7 +86,7 @@ static void show_faulting_vma(unsigned long address)
struct vm_area_struct *vma;
struct mm_struct *active_mm = current->active_mm;
- /* can't use print_vma_addr() yet as it doesn't check for
+ /* can't use %pv yet as it doesn't check for
* non-inclusive vma
*/
mmap_read_lock(active_mm);
diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
index 13ebd5ca2070..294ed81f67d8 100644
--- a/arch/arm64/kernel/traps.c
+++ b/arch/arm64/kernel/traps.c
@@ -224,13 +224,15 @@ static void arm64_show_signal(int signo, const char *str)
!__ratelimit())
return;
- pr_info("%s[%d]: unhandled exception: ", tsk->comm, task_pid_nr(tsk));
- if (esr)
- pr_cont("%s, ESR 0x%08x, ", esr_get_class_string(esr), esr);
-
- pr_cont("%s", str);
- print_vma_addr(KERN_CONT " in ", regs->pc);
- pr_cont("\n");
+ if (esr) {
+ pr_info("%s[%d]: unhandled exception: %s, ESR 0x%08x, %s in
%pv\n",
+ tsk->comm, task_pid_nr(tsk),
+ esr_get_class_string(esr), esr,
+ str, (void *)regs->pc);
+ } else {
+ pr_info("%s[%d]: unhandled exception: %s in %pv\n",
+ tsk->comm, task_pid_nr(tsk), str, (void *)regs->pc);
+ }
__show_regs(regs);
}
diff --git a/arch/csky/kernel/traps.c b/arch/csky/kernel/traps.c
index 959a917c989d..bdf1577237df 100644
--- a/arch/csky/kernel/traps.c
+++ b/arch/csky/kernel/traps.c
@@ -118,12 +118,11 @@ void do_trap(struct pt_regs *regs, int signo, int code,
unsigned long addr)
{
struct task_struct *tsk = current;
- if (show_unhandled_signals && unhandled_signal(tsk, signo)
- && printk_ratelimit()) {
- pr_info("%s[%d]: unhandled signal %d code 0x%x at 0x%08lx",
- tsk->comm, task_pid_nr(tsk), signo, code, addr);
- print_vma_addr(KERN_CONT " in ", instruction_pointer(regs));
- pr_cont("\n");
+ if (show_unhandled_signals && unhandled_signal(tsk, signo) &&
+ printk_ratelimit()) {
+ pr_info("%s[%d]: unhandled signal %d code 0x%x at 0x%08lx in
%pv\n",
+ tsk->comm, task_pid_nr(tsk), signo, code, addr,
+ (void *)instruction_pointer(regs));
show_regs(regs);
}
diff --git
Re: [RFC PATCH] vsprintf: Add %pv extension replacement for print_vma_addr
On Fri, 14 Aug 2020 10:53:03 -0700
Joe Perches wrote:
> There is a print_vma_addr function used several places
> that requires KERN_CONT use.
>
> Add a %pv mechanism to avoid the need for KERN_CONT.
>
> An example conversion is arch/x86/kernel/signal.c
>
> from:
> if (show_unhandled_signals && printk_ratelimit()) {
> printk("%s"
> "%s[%d] bad frame in %s frame:%p ip:%lx sp:%lx orax:%lx",
> task_pid_nr(current) > 1 ? KERN_INFO : KERN_EMERG,
> me->comm, me->pid, where, frame,
> regs->ip, regs->sp, regs->orig_ax);
> print_vma_addr(KERN_CONT " in ", regs->ip);
> pr_cont("\n");
> to:
> printk("%s"
> "%s[%d] bad frame in %s frame:%p ip:%lx sp:%lx orax:%lx
> in %pv\n",
> task_pid_nr(current) > 1 ? KERN_INFO : KERN_EMERG,
> me->comm, me->pid, where, frame,
> regs->ip, regs->sp, regs->orig_ax, (void *)regs->ip);
> ---
> lib/vsprintf.c | 52
> 1 file changed, 52 insertions(+)
>
> diff --git a/lib/vsprintf.c b/lib/vsprintf.c
> index c155769559ab..654402c43f8d 100644
> --- a/lib/vsprintf.c
> +++ b/lib/vsprintf.c
> @@ -995,6 +995,12 @@ static const struct printf_spec default_dec_spec = {
> .precision = -1,
> };
>
> +static const struct printf_spec default_hex_spec = {
> + .base = 16,
> + .precision = -1,
> + .flags = SMALL,
> +};
> +
> static const struct printf_spec default_dec02_spec = {
> .base = 10,
> .field_width = 2,
> @@ -2089,6 +2095,50 @@ char *fwnode_string(char *buf, char *end, struct
> fwnode_handle *fwnode,
> return widen_string(buf, buf - buf_start, end, spec);
> }
>
> +static noinline_for_stack
> +char *vma_addr(char *buf, char *end, void *ip,
> +struct printf_spec spec, const char *fmt)
> +{
> +#ifdef CONFIG_MMU
> + struct mm_struct *mm = current->mm;
> + struct vm_area_struct *vma;
> +
> + /*
> + * we might be running from an atomic context so we cannot sleep
> + */
> + if (!mmap_read_trylock(mm))
> + return buf;
> +
> + vma = find_vma(mm, (unsigned long)ip);
> + if (vma && vma->vm_file) {
> + char *p;
> + struct file *f = vma->vm_file;
> + char *page = (char *)__get_free_page(GFP_NOWAIT);
> +
> + if (page) {
> + p = file_path(f, page, PAGE_SIZE);
> + if (IS_ERR(p))
> + p = "?";
> + buf = string(buf, end, kbasename(p), default_str_spec);
> + buf = string_nocheck(buf, end, "[", default_str_spec);
> + buf = pointer_string(buf, end,
> + (void *)vma->vm_start,
> + default_hex_spec);
> + buf = string_nocheck(buf, end, "+", default_str_spec);
> + buf = pointer_string(buf, end,
> + (void *)(vma->vm_end -
> vma->vm_start),
> + default_hex_spec);
> + buf = string_nocheck(buf, end, "]", default_str_spec);
> + free_page((unsigned long)page);
> + }
> + }
> + mmap_read_unlock(mm);
> +#else
> + buf = string_nocheck(buf, end, "CONFIG_MMU=n", default_str_spec);
> +#endif
I'm fine with all his, but I feel more comfortable if this patch
created a single copy of the code. Perhaps add:
diff --git a/mm/memory.c b/mm/memory.c
index 87ec87cdc1ff..795a4db4d83d 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -4771,32 +4771,7 @@ EXPORT_SYMBOL_GPL(access_process_vm);
*/
void print_vma_addr(char *prefix, unsigned long ip)
{
- struct mm_struct *mm = current->mm;
- struct vm_area_struct *vma;
-
- /*
-* we might be running from an atomic context so we cannot sleep
-*/
- if (!mmap_read_trylock(mm))
- return;
-
- vma = find_vma(mm, ip);
- if (vma && vma->vm_file) {
- struct file *f = vma->vm_file;
- char *buf = (char *)__get_free_page(GFP_NOWAIT);
- if (buf) {
- char *p;
-
- p = file_path(f, buf, PAGE_SIZE);
- if (IS_ERR(p))
- p = "?";
- printk("%s%s[%lx+%lx]", prefix, kbasename(p),
- vma->vm_start,
- vma->vm_end - vma->vm_start);
- free_page((unsigned long)buf);
- }
- }
- mmap_read_unlock(mm);
+ printk("%s%pv", prefix, ip);
}
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
And of course this would need for documentation.
--
[RFC PATCH] vsprintf: Add %pv extension replacement for print_vma_addr
There is a print_vma_addr function used several places
that requires KERN_CONT use.
Add a %pv mechanism to avoid the need for KERN_CONT.
An example conversion is arch/x86/kernel/signal.c
from:
if (show_unhandled_signals && printk_ratelimit()) {
printk("%s"
"%s[%d] bad frame in %s frame:%p ip:%lx sp:%lx orax:%lx",
task_pid_nr(current) > 1 ? KERN_INFO : KERN_EMERG,
me->comm, me->pid, where, frame,
regs->ip, regs->sp, regs->orig_ax);
print_vma_addr(KERN_CONT " in ", regs->ip);
pr_cont("\n");
to:
printk("%s"
"%s[%d] bad frame in %s frame:%p ip:%lx sp:%lx orax:%lx
in %pv\n",
task_pid_nr(current) > 1 ? KERN_INFO : KERN_EMERG,
me->comm, me->pid, where, frame,
regs->ip, regs->sp, regs->orig_ax, (void *)regs->ip);
---
lib/vsprintf.c | 52
1 file changed, 52 insertions(+)
diff --git a/lib/vsprintf.c b/lib/vsprintf.c
index c155769559ab..654402c43f8d 100644
--- a/lib/vsprintf.c
+++ b/lib/vsprintf.c
@@ -995,6 +995,12 @@ static const struct printf_spec default_dec_spec = {
.precision = -1,
};
+static const struct printf_spec default_hex_spec = {
+ .base = 16,
+ .precision = -1,
+ .flags = SMALL,
+};
+
static const struct printf_spec default_dec02_spec = {
.base = 10,
.field_width = 2,
@@ -2089,6 +2095,50 @@ char *fwnode_string(char *buf, char *end, struct
fwnode_handle *fwnode,
return widen_string(buf, buf - buf_start, end, spec);
}
+static noinline_for_stack
+char *vma_addr(char *buf, char *end, void *ip,
+ struct printf_spec spec, const char *fmt)
+{
+#ifdef CONFIG_MMU
+ struct mm_struct *mm = current->mm;
+ struct vm_area_struct *vma;
+
+ /*
+* we might be running from an atomic context so we cannot sleep
+*/
+ if (!mmap_read_trylock(mm))
+ return buf;
+
+ vma = find_vma(mm, (unsigned long)ip);
+ if (vma && vma->vm_file) {
+ char *p;
+ struct file *f = vma->vm_file;
+ char *page = (char *)__get_free_page(GFP_NOWAIT);
+
+ if (page) {
+ p = file_path(f, page, PAGE_SIZE);
+ if (IS_ERR(p))
+ p = "?";
+ buf = string(buf, end, kbasename(p), default_str_spec);
+ buf = string_nocheck(buf, end, "[", default_str_spec);
+ buf = pointer_string(buf, end,
+(void *)vma->vm_start,
+default_hex_spec);
+ buf = string_nocheck(buf, end, "+", default_str_spec);
+ buf = pointer_string(buf, end,
+(void *)(vma->vm_end -
vma->vm_start),
+default_hex_spec);
+ buf = string_nocheck(buf, end, "]", default_str_spec);
+ free_page((unsigned long)page);
+ }
+ }
+ mmap_read_unlock(mm);
+#else
+ buf = string_nocheck(buf, end, "CONFIG_MMU=n", default_str_spec);
+#endif
+ return buf;
+}
+
/*
* Show a '%p' thing. A kernel extension is that the '%p' is followed
* by an extra set of alphanumeric characters that are extended format
@@ -2254,6 +2304,8 @@ char *pointer(const char *fmt, char *buf, char *end, void
*ptr,
return uuid_string(buf, end, ptr, spec, fmt);
case 'V':
return va_format(buf, end, ptr, spec, fmt);
+ case 'v':
+ return vma_addr(buf, end, ptr, spec, fmt);
case 'K':
return restricted_pointer(buf, end, ptr, spec);
case 'N':
