coreboot-kgpe-d16/payloads/libpayload/libc/printf.c

824 lines
21 KiB
C
Raw Normal View History

/*
* This file is part of the libpayload project.
*
* It has originally been taken from the HelenOS project
* (http://www.helenos.eu), and slightly modified for our purposes.
*
* Copyright (C) 2001-2004 Jakub Jermar
* Copyright (C) 2006 Josef Cejka
* Copyright (C) 2008 Uwe Hermann <uwe@hermann-uwe.de>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* - Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* - Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* - The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <libpayload.h>
#include <ctype.h>
static struct _FILE {
} _stdout, _stdin, _stderr;
FILE *stdout = &_stdout;
FILE *stdin = &_stdin;
FILE *stderr = &_stderr;
/** Structure for specifying output methods for different printf clones. */
struct printf_spec {
/* Output function, returns count of printed characters or EOF. */
int (*write) (const char *, size_t, void *);
/* Support data - output stream specification, its state, locks, ... */
void *data;
};
/** Show prefixes 0x or 0. */
#define __PRINTF_FLAG_PREFIX 0x00000001
/** Signed / unsigned number. */
#define __PRINTF_FLAG_SIGNED 0x00000002
/** Print leading zeroes. */
#define __PRINTF_FLAG_ZEROPADDED 0x00000004
/** Align to left. */
#define __PRINTF_FLAG_LEFTALIGNED 0x00000010
/** Always show + sign. */
#define __PRINTF_FLAG_SHOWPLUS 0x00000020
/** Print space instead of plus. */
#define __PRINTF_FLAG_SPACESIGN 0x00000040
/** Show big characters. */
#define __PRINTF_FLAG_BIGCHARS 0x00000080
/** Number has - sign. */
#define __PRINTF_FLAG_NEGATIVE 0x00000100
/**
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
* Buffer big enough for 64-bit number printed in base 2, sign, and prefix.
* Add some more to support sane amounts of zero-padding.
*/
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
#define PRINT_BUFFER_SIZE (64 + 1 + 2 + 13)
/** Enumeration of possible arguments types. */
typedef enum {
PrintfQualifierByte = 0,
PrintfQualifierShort,
PrintfQualifierInt,
PrintfQualifierLong,
PrintfQualifierLongLong,
PrintfQualifierPointer,
} qualifier_t;
static const char digits_small[] = "0123456789abcdef";
static const char digits_big[] = "0123456789ABCDEF";
/**
* Print one or more characters without adding newline.
*
* @param buf Buffer of >= count bytesi size. NULL pointer is not allowed!
* @param count Number of characters to print.
* @param ps Output method and its data.
* @return Number of characters printed.
*/
static int printf_putnchars(const char *buf, size_t count,
struct printf_spec *ps)
{
return ps->write(buf, count, ps->data);
}
/**
* Print a string without adding a newline.
*
* @param str String to print.
* @param ps Write function specification and support data.
* @return Number of characters printed.
*/
static inline int printf_putstr(const char *str, struct printf_spec *ps)
{
return printf_putnchars(str, strlen(str), ps);
}
/**
* Print one character.
*
* @param c Character to be printed.
* @param ps Output method.
* @return Number of characters printed.
*/
static int printf_putchar(int c, struct printf_spec *ps)
{
char ch = c;
return ps->write(&ch, 1, ps->data);
}
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
/* Print spaces for padding. Ignores negative counts. */
static int print_spaces(int count, struct printf_spec *ps)
{
int tmp, ret;
char buffer[PRINT_BUFFER_SIZE];
if (count <= 0)
return 0;
memset(buffer, ' ', MIN(PRINT_BUFFER_SIZE, count));
for (tmp = count; tmp > PRINT_BUFFER_SIZE; tmp -= PRINT_BUFFER_SIZE)
if ((ret = printf_putnchars(buffer, PRINT_BUFFER_SIZE, ps)) < 0)
return ret;
if ((ret = printf_putnchars(buffer, tmp, ps)) < 0)
return ret;
return count;
}
/**
* Print one formatted character.
*
* @param c Character to print.
* @param width Width modifier.
* @param flags Flags that change the way the character is printed.
* @param ps Output methods spec for different printf clones.
* @return Number of characters printed, negative value on failure.
*/
static int print_char(char c, int width, uint64_t flags, struct printf_spec *ps)
{
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
int retval;
int counter = 1;
if (!(flags & __PRINTF_FLAG_LEFTALIGNED)) {
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
if ((retval = print_spaces(width - 1, ps)) < 0)
return retval;
else
counter += retval;
}
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
if ((retval = printf_putchar(c, ps)) < 0)
return retval;
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
if (flags & __PRINTF_FLAG_LEFTALIGNED) {
if ((retval = print_spaces(width - 1, ps)) < 0)
return retval;
else
counter += retval;
}
return counter;
}
/**
* Print string.
*
* @param s String to be printed.
* @param width Width modifier.
* @param precision Precision modifier.
* @param flags Flags that modify the way the string is printed.
* @param ps Output methods spec for different printf clones.
* @return Number of characters printed, negative value on failure.
*/
/** Structure for specifying output methods for different printf clones. */
static int print_string(char *s, int width, unsigned int precision,
uint64_t flags, struct printf_spec *ps)
{
int counter = 0, retval;
size_t size;
if (s == NULL)
return printf_putstr("(NULL)", ps);
size = strlen(s);
/* Print leading spaces. */
if (precision == 0)
precision = size;
width -= precision;
if (!(flags & __PRINTF_FLAG_LEFTALIGNED)) {
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
if ((retval = print_spaces(width, ps)) < 0)
return retval;
else
counter += retval;
}
if ((retval = printf_putnchars(s, MIN(size, precision), ps)) < 0)
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
return retval;
counter += retval;
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
if (flags & __PRINTF_FLAG_LEFTALIGNED) {
if ((retval = print_spaces(width, ps)) < 0)
return retval;
else
counter += retval;
}
return counter;
}
/**
* Print a number in a given base.
*
* Print significant digits of a number in given base.
*
* @param num Number to print.
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
* @param width Width modifier.
* @param precision Precision modifier.
* @param base Base to print the number in (must be between 2 and 16).
* @param flags Flags that modify the way the number is printed.
* @param ps Output methods spec for different printf clones.
* @return Number of characters printed.
*/
static int print_number(uint64_t num, int width, int precision, int base,
uint64_t flags, struct printf_spec *ps)
{
const char *digits = digits_small;
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
char d[PRINT_BUFFER_SIZE];
char *ptr = &d[PRINT_BUFFER_SIZE];
int size = 0; /* Size of the string in ptr */
int counter = 0; /* Amount of actually printed bytes. */
char sgn;
int retval;
if (flags & __PRINTF_FLAG_BIGCHARS)
digits = digits_big;
if (num == 0) {
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
*--ptr = '0';
size++;
} else {
do {
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
*--ptr = digits[num % base];
size++;
} while (num /= base);
}
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
/* Both precision and LEFTALIGNED overrule ZEROPADDED. */
if ((flags & __PRINTF_FLAG_LEFTALIGNED) || precision)
flags &= ~__PRINTF_FLAG_ZEROPADDED;
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
/* Fix precision now since it doesn't count prefixes/signs. */
precision -= size;
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
/* Reserve size for prefixes/signs before filling up padding. */
sgn = 0;
if (flags & __PRINTF_FLAG_SIGNED) {
if (flags & __PRINTF_FLAG_NEGATIVE) {
sgn = '-';
size++;
} else if (flags & __PRINTF_FLAG_SHOWPLUS) {
sgn = '+';
size++;
} else if (flags & __PRINTF_FLAG_SPACESIGN) {
sgn = ' ';
size++;
}
}
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
if (flags & __PRINTF_FLAG_PREFIX) {
switch (base) {
case 2: /* Binary formating is not standard, but useful. */
size += 2;
break;
case 8:
size++;
break;
case 16:
size += 2;
break;
}
}
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
/* If this is still set we didn't have a precision, so repurpose it */
if (flags & __PRINTF_FLAG_ZEROPADDED)
precision = width - size;
/* Pad smaller numbers with 0 (larger numbers lead to precision < 0). */
if (precision > 0) {
precision = MIN(precision, PRINT_BUFFER_SIZE - size);
ptr -= precision;
size += precision;
memset(ptr, '0', precision);
}
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
/* Add sign and prefix (we adjusted size for this beforehand). */
if (flags & __PRINTF_FLAG_PREFIX) {
switch (base) {
case 2: /* Binary formating is not standard, but useful. */
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
*--ptr = (flags & __PRINTF_FLAG_BIGCHARS) ? 'B' : 'b';
*--ptr = '0';
break;
case 8:
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
*--ptr = '0';
break;
case 16:
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
*--ptr = (flags & __PRINTF_FLAG_BIGCHARS) ? 'X' : 'x';
*--ptr = '0';
break;
}
}
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
if (sgn)
*--ptr = sgn;
/* Pad with spaces up to width, try to avoid extra putnchar if we can */
width -= size;
if (width > 0 && !(flags & __PRINTF_FLAG_LEFTALIGNED)) {
int tmp = MIN(width, PRINT_BUFFER_SIZE - size);
ptr -= tmp;
size += tmp;
memset(ptr, ' ', tmp);
if ((retval = print_spaces(width - tmp, ps)) < 0)
return retval;
else
counter += retval;
}
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
/* Now print the whole thing at once. */
if ((retval = printf_putnchars(ptr, size, ps)) < 0)
return retval;
counter += retval;
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
/* Edge case: left-aligned with width (should be rare). */
if (flags & __PRINTF_FLAG_LEFTALIGNED) {
if ((retval = print_spaces(width, ps)) < 0)
return retval;
else
counter += retval;
}
return counter;
}
/**
* Print formatted string.
*
* Print string formatted according to the fmt parameter and variadic arguments.
* Each formatting directive must have the following form:
*
* \% [ FLAGS ] [ WIDTH ] [ .PRECISION ] [ TYPE ] CONVERSION
*
* FLAGS:@n
* - "#" Force to print prefix.For \%o conversion, the prefix is 0, for
* \%x and \%X prefixes are 0x and 0X and for conversion \%b the
* prefix is 0b.
*
* - "-" Align to left.
*
* - "+" Print positive sign just as negative.
*
* - " " If the printed number is positive and "+" flag is not set,
* print space in place of sign.
*
* - "0" Print 0 as padding instead of spaces. Zeroes are placed between
* sign and the rest of the number. This flag is ignored if "-"
* flag is specified.
*
* WIDTH:@n
* - Specify the minimal width of a printed argument. If it is bigger,
* width is ignored. If width is specified with a "*" character instead of
* number, width is taken from parameter list. And integer parameter is
* expected before parameter for processed conversion specification. If
* this value is negative its absolute value is taken and the "-" flag is
* set.
*
* PRECISION:@n
* - Value precision. For numbers it specifies minimum valid numbers.
* Smaller numbers are printed with leading zeroes. Bigger numbers are not
* affected. Strings with more than precision characters are cut off. Just
* as with width, an "*" can be used used instead of a number. An integer
* value is then expected in parameters. When both width and precision are
* specified using "*", the first parameter is used for width and the
* second one for precision.
*
* TYPE:@n
* - "hh" Signed or unsigned char.@n
* - "h" Signed or unsigned short.@n
* - "" Signed or unsigned int (default value).@n
* - "l" Signed or unsigned long int.@n
* - "ll" Signed or unsigned long long int.@n
*
*
* CONVERSION:@n
* - % Print percentile character itself.
*
* - c Print single character.
*
* - s Print zero terminated string. If a NULL value is passed as
* value, "(NULL)" is printed instead.
*
* - P, p Print value of a pointer. Void * value is expected and it is
* printed in hexadecimal notation with prefix (as with \%#X / \%#x
* for 32-bit or \%#X / \%#x for 64-bit long pointers).
*
* - b Print value as unsigned binary number. Prefix is not printed by
* default. (Nonstandard extension.)
*
* - o Print value as unsigned octal number. Prefix is not printed by
* default.
*
* - d, i Print signed decimal number. There is no difference between d
* and i conversion.
*
* - u Print unsigned decimal number.
*
* - X, x Print hexadecimal number with upper- or lower-case. Prefix is
* not printed by default.
*
* All other characters from fmt except the formatting directives are printed in
* verbatim.
*
* @param fmt Formatting NULL terminated string.
* @param ps TODO.
* @param ap TODO.
* @return Number of characters printed, negative value on failure.
*/
static int printf_core(const char *fmt, struct printf_spec *ps, va_list ap)
{
int i = 0; /* Index of the currently processed char from fmt */
int j = 0; /* Index to the first not printed nonformating character */
int end;
int counter; /* Counter of printed characters */
int retval; /* Used to store return values from called functions */
char c;
qualifier_t qualifier; /* Type of argument */
int base; /* Base in which a numeric parameter will be printed */
uint64_t number; /* Argument value */
size_t size; /* Byte size of integer parameter */
int width, precision;
uint64_t flags;
counter = 0;
while ((c = fmt[i])) {
/* Control character. */
if (c == '%') {
/* Print common characters if any processed. */
if (i > j) {
if ((retval = printf_putnchars(&fmt[j],
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
(size_t) (i - j), ps)) < 0)
return retval;
counter += retval;
}
j = i;
/* Parse modifiers. */
flags = 0;
end = 0;
do {
++i;
switch (c = fmt[i]) {
case '#':
flags |= __PRINTF_FLAG_PREFIX;
break;
case '-':
flags |= __PRINTF_FLAG_LEFTALIGNED;
break;
case '+':
flags |= __PRINTF_FLAG_SHOWPLUS;
break;
case ' ':
flags |= __PRINTF_FLAG_SPACESIGN;
break;
case '0':
flags |= __PRINTF_FLAG_ZEROPADDED;
break;
default:
end = 1;
};
} while (end == 0);
/* Width & '*' operator. */
width = 0;
if (isdigit(fmt[i])) {
while (isdigit(fmt[i])) {
width *= 10;
width += fmt[i++] - '0';
}
} else if (fmt[i] == '*') {
/* Get width value from argument list. */
i++;
width = (int)va_arg(ap, int);
if (width < 0) {
/* Negative width sets '-' flag. */
width *= -1;
flags |= __PRINTF_FLAG_LEFTALIGNED;
}
}
/* Precision and '*' operator. */
precision = 0;
if (fmt[i] == '.') {
++i;
if (isdigit(fmt[i])) {
while (isdigit(fmt[i])) {
precision *= 10;
precision += fmt[i++] - '0';
}
} else if (fmt[i] == '*') {
/* Get precision from argument list. */
i++;
precision = (int)va_arg(ap, int);
/* Ignore negative precision. */
if (precision < 0)
precision = 0;
}
}
switch (fmt[i++]) {
/** @todo unimplemented qualifiers:
* t ptrdiff_t - ISO C 99
*/
case 'h': /* char or short */
qualifier = PrintfQualifierShort;
if (fmt[i] == 'h') {
i++;
qualifier = PrintfQualifierByte;
}
break;
case 'z': /* size_t or ssize_t */
qualifier = PrintfQualifierLong;
break;
case 'l': /* long or long long */
qualifier = PrintfQualifierLong;
if (fmt[i] == 'l') {
i++;
qualifier = PrintfQualifierLongLong;
}
break;
default:
/* default type */
qualifier = PrintfQualifierInt;
--i;
}
base = 10;
switch (c = fmt[i]) {
/* String and character conversions */
case 's':
if ((retval = print_string(va_arg(ap, char *),
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
width, precision, flags, ps)) < 0)
return retval;
counter += retval;
j = i + 1;
goto next_char;
case 'c':
c = va_arg(ap, unsigned int);
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
if ((retval = print_char(c, width, flags, ps)) < 0)
return retval;
counter += retval;
j = i + 1;
goto next_char;
/* Integer values */
case 'P': /* pointer */
flags |= __PRINTF_FLAG_BIGCHARS;
case 'p':
flags |= __PRINTF_FLAG_PREFIX;
base = 16;
qualifier = PrintfQualifierPointer;
break;
case 'b':
base = 2;
break;
case 'o':
base = 8;
break;
case 'd':
case 'i':
flags |= __PRINTF_FLAG_SIGNED;
case 'u':
break;
case 'X':
flags |= __PRINTF_FLAG_BIGCHARS;
case 'x':
base = 16;
break;
case '%': /* percentile itself */
j = i;
goto next_char;
default: /* Bad formatting */
/*
* Unknown format. Now, j is the index of '%'
* so we will print whole bad format sequence.
*/
goto next_char;
}
/* Print integers. */
/* Print number. */
switch (qualifier) {
case PrintfQualifierByte:
size = sizeof(unsigned char);
number = (uint64_t) va_arg(ap, unsigned int);
break;
case PrintfQualifierShort:
size = sizeof(unsigned short);
number = (uint64_t) va_arg(ap, unsigned int);
break;
case PrintfQualifierInt:
size = sizeof(unsigned int);
number = (uint64_t) va_arg(ap, unsigned int);
break;
case PrintfQualifierLong:
size = sizeof(unsigned long);
number = (uint64_t) va_arg(ap, unsigned long);
break;
case PrintfQualifierLongLong:
size = sizeof(unsigned long long);
number = (uint64_t) va_arg(ap, unsigned long long);
break;
case PrintfQualifierPointer:
size = sizeof(void *);
number = (uint64_t) (unsigned long)va_arg(ap, void *);
break;
}
if (flags & __PRINTF_FLAG_SIGNED) {
if (number & (0x1ULL << (size * 8 - 1))) {
flags |= __PRINTF_FLAG_NEGATIVE;
if (size == sizeof(uint64_t)) {
number = -((int64_t) number);
} else {
number = ~number;
number &= ~(0xFFFFFFFFFFFFFFFFll << (size * 8));
number++;
}
}
}
if ((retval = print_number(number, width, precision,
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
base, flags, ps)) < 0)
return retval;
counter += retval;
j = i + 1;
}
next_char:
++i;
}
if (i > j) {
if ((retval = printf_putnchars(&fmt[j],
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
(u64) (i - j), ps)) < 0)
return retval;
counter += retval;
}
return counter;
}
int snprintf(char *str, size_t size, const char *fmt, ...)
{
int ret;
va_list args;
va_start(args, fmt);
ret = vsnprintf(str, size, fmt, args);
va_end(args);
return ret;
}
int sprintf(char *str, const char *fmt, ...)
{
int ret;
va_list args;
va_start(args, fmt);
ret = vsprintf(str, fmt, args);
va_end(args);
return ret;
}
int fprintf(FILE *file, const char *fmt, ...)
{
int ret;
if ((file == stdout) || (file == stderr)) {
va_list args;
va_start(args, fmt);
ret = vprintf(fmt, args);
va_end(args);
return ret;
}
return -1;
}
struct vsnprintf_data {
size_t size; /* Total space for string */
size_t len; /* Count of currently used characters */
char *string; /* Destination string */
};
/**
* Write string to given buffer.
*
* Write at most data->size characters including trailing zero. According to
* C99, snprintf() has to return number of characters that would have been
* written if enough space had been available. Hence the return value is not
* number of really printed characters but size of the input string.
* Number of really used characters is stored in data->len.
*
* @param str Source string to print.
* @param count Size of source string.
* @param _data Structure with destination string, counter of used space
* and total string size.
* @return Number of characters to print (not characters really printed!).
*/
static int vsnprintf_write(const char *str, size_t count, void *_data)
{
struct vsnprintf_data *data = _data;
size_t i;
i = data->size - data->len;
if (i == 0)
return count;
/* We have only one free byte left in buffer => write trailing zero. */
if (i == 1) {
data->string[data->size - 1] = 0;
data->len = data->size;
return count;
}
/*
* We have not enough space for whole string with the trailing
* zero => print only a part of string.
*/
if (i <= count) {
memcpy((void *)(data->string + data->len), (void *)str, i - 1);
data->string[data->size - 1] = 0;
data->len = data->size;
return count;
}
/* Buffer is big enough to print whole string. */
memcpy((void *)(data->string + data->len), (void *)str, count);
data->len += count;
/*
* Put trailing zero at end, but not count it into data->len so
* it could be rewritten next time.
*/
data->string[data->len] = 0;
return count;
}
int vsnprintf(char *str, size_t size, const char *fmt, va_list ap)
{
struct vsnprintf_data data = { size, 0, str };
struct printf_spec ps = { vsnprintf_write, &data };
/* Print 0 at end of string - fix case that nothing will be printed. */
if (size > 0)
str[0] = 0;
/* vsnprintf_write() ensures that str will be terminated by zero. */
return printf_core(fmt, &ps, ap);
}
int vsprintf(char *str, const char *fmt, va_list ap)
{
return vsnprintf(str, (size_t) - 1, fmt, ap);
}
int printf(const char *fmt, ...)
{
int ret;
va_list args;
va_start(args, fmt);
ret = vprintf(fmt, args);
va_end(args);
return ret;
}
static int vprintf_write(const char *str, size_t count, void *unused)
{
libpayload: console: Allow output drivers to print whole strings at once The console output driver framework in libpayload is currently built on the putchar primitive, meaning that every driver's function gets called one character at a time. This becomes an issue when we add drivers that could output multiple characters at a time, but have a high constant overhead per invocation (such as the planned GDB stub, which needs to wrap a special frame around output strings and wait for an acknowledgement from the server). This patch adds a new 'write' function pointer to the console_output_driver structure as an alternative to 'putchar'. Output drivers need to provide at least one of the two ('write' is preferred if available). The CBMEM console driver is ported as a proof of concept (since it's our most performace-critical driver and should in theory benefit the most from less function pointer invocations, although it's probably still negligible compared to the big sprawling mess that is printf()). Even with this fix, the problem remains that printf() was written with the putchar primitive in mind. Even though normal text already contains an optimization to allow multiple characters at a time, almost all formatting directives cause their output (including things like padding whitespace) to be putchar()ed one character at a time. Therefore, this patch reworks parts of the output code (especially number printing) to all but remove that inefficiency (directives still invoke an extra write() call, but at least not one per character). Since I'm touching printf() core code anyway, I also tried to salvage what I could from that weird, broken "return negative on error" code path (not that any of our current output drivers can trigger it anyway). A final consequence of this patch is that the responsibility to prepend line feeds with carriage returns is moved into the output driver implementations. Doing this only makes sense for drivers with explicit cursor position control (i.e. serial or video), and things like the CBMEM console that appears like a normal file to the system really have no business containing carriage returns (we don't want people to accidentally associate us with Windows, now, do we?). BUG=chrome-os-partner:18390 TEST=Made sure video and CBMEM console still look good, tried printf() with as many weird edge-case strings as I could find and compared serial output as well as sprintf() return value. Original-Change-Id: Ie05ae489332a0103461620f5348774b6d4afd91a Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/196384 Original-Reviewed-by: Hung-Te Lin <hungte@chromium.org> Original-Reviewed-by: David Hendricks <dhendrix@chromium.org> (cherry picked from commit ab1ef0c07736fe1aa3e0baaf02d258731e6856c0) Signed-off-by: Marc Jones <marc.jones@se-eng.com> Change-Id: I78f5aedf6d0c3665924995cdab691ee0162de404 Reviewed-on: http://review.coreboot.org/7880 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2014-04-18 05:00:20 +02:00
console_write(str, count);
return count;
}
int vprintf(const char *fmt, va_list ap)
{
struct printf_spec ps = { vprintf_write, NULL };
return printf_core(fmt, &ps, ap);
}