Re: 答复: [External Mail]Re: [PATCH v4] Fix: move huge compressed obj from page to zspage

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 2023/10/13 14:03, 陈冠有 wrote:
> Hi Kazu
> 
> Thank you for your advice.
> 
> 
>> Does this mean that the current "memset(buf, entry, len)" is a bug,
>> because entry is ulong, right?  If so, I agree.
> 
> 
> Yes.
> 
> 
> Subject: [PATCH] Fix: "rd" command to read zram buf on Linux 5.17 and later
> 
> Kernel commit a41ec880aa7b ("zsmalloc: move huge compressed obj from page to zspage"),
> which is contained in Linux 5.17 and later kernels. Use zspage_5_17 to be compatible with later versions.
> 
> Kernel commit ffedd09fa9b06 ("zsmalloc: Stop using slab fields in struct page"),
> page->index and page->freelist are at the same offset in struct page.
> 
> Kernel commit f635725c3905e ("zram: do not waste zram_table_entry flags bits"),
> which is contained in Linux 6.1 and later kernels. let ZRAM_FLAG_SHIFT follow ZRAM_LOCK.
> 
> Before:
> crash> mod -s zram zram.ko
>       MODULE       NAME                                          BASE           SIZE  OBJECT FILE
> ffffffde224db800  zram                                    ffffffde224d2000    57344  zram.ko
> crash> mod -s zsmalloc zsmalloc.ko
>       MODULE       NAME                                          BASE           SIZE  OBJECT FILE
> ffffffde224c5180  zsmalloc                                ffffffde224bf000    40960  zsmalloc.ko
> crash> rd 0x13d89fb0
> rd: zspage magic incorrect: b0
> 
> After:
> crash> rd 0x13d89fb0
>          13d89fb0:  c2b54f7170883b20                     ;.pqO..
> 
> Link: https://lkml.kernel.org/r/20211115185909.3949505-6-minchan@xxxxxxxxxx
> Link: https://lkml.kernel.org/r/20220912152744.527438-1-senozhatsky@xxxxxxxxxxxx
> 
> Signed-off-by: chenguanyou <chenguanyou@xxxxxxxxxx>
> ---
>   defs.h     | 26 +++++++++++++++++++++
>   diskdump.c | 66 ++++++++++++++++++++++++++++++++++++++++++++----------
>   2 files changed, 80 insertions(+), 12 deletions(-)
> 
> diff --git a/defs.h b/defs.h
> index 96a7a2a..2038351 100644
> --- a/defs.h
> +++ b/defs.h
> @@ -2225,6 +2225,7 @@ struct offset_table {                    /* stash of commonly-used offsets */
>      long module_memory_base;
>      long module_memory_size;
>      long irq_data_irq;
> +   long zspage_huge;
>   };
> 
>   struct size_table {         /* stash of commonly-used sizes */
> @@ -7210,6 +7211,19 @@ ulong try_zram_decompress(ulonglong pte_val, unsigned char *buf, ulong len, ulon
>   #define SECTORS_PER_PAGE        (1 << SECTORS_PER_PAGE_SHIFT)
>   #define ZRAM_FLAG_SHIFT         (1<<24)
>   #define ZRAM_FLAG_SAME_BIT      (1<<25)
> +
> +struct zram_pageflags {
> +    long ZRAM_LOCK;
> +    long ZRAM_SAME;
> +};
> +
> +/*
> + *  diskdump.c
> + */
> +extern struct zram_pageflags zram_pageflags;
> +#define ZRAM_PAGEFLAG_INIT(X, Y) (zram_pageflags.X = Y)
> +#define ZRAM_PAGEFLAG_VALUE(X) (zram_pageflags.X)

Thank you for the update.

I don't think these macros are necessary, so simplified the patch on my 
end.  With rethinking the names, kept "ZRAM_FLAG_SHIFT" and 
"ZRAM_FLAG_SAME_BIT" as they are, and made some tweaks.

Guanyou and Lianbo, is this attached patch OK?

I tested this on Linux 6.1.

crash> vtop 00007fb60be00000
VIRTUAL     PHYSICAL
7fb60be00000  (not mapped)

    PGD: 108e807f8 => 1107c4067
    PUD: 1107c46c0 => 106ec9067
    PMD: 106ec92f8 => 1060f2067
    PTE: 1060f2000 => 7fffffffdfffe0a

       PTE           SWAP     OFFSET
7fffffffdfffe0a  /dev/zram0   65536

       VMA           START       END     FLAGS FILE
ffff935206cda980 7fb60bdfe000 7fb64be00000 8100073

SWAP: /dev/zram0  OFFSET: 65536

crash> rd 00007fb60be00000
     7fb60be00000:  0000000000001000                    ........

(Note that it looks like Linux 6.2 and later have other changes related 
to zram, so this patch supports up to 6.1.  We can fix them later.)

Thanks,
Kazu


> +
>   struct zspage {
>       struct {
>           unsigned int fullness : 2;
> @@ -7221,6 +7235,18 @@ struct zspage {
>       unsigned int freeobj;
>   };
> 
> +struct zspage_5_17 {
> +    struct {
> +        unsigned int huge : 1;
> +        unsigned int fullness : 2;
> +        unsigned int class : 9;
> +        unsigned int isolated : 3;
> +        unsigned int magic : 8;
> +    };
> +    unsigned int inuse;
> +    unsigned int freeobj;
> +};
> +
>   /*
>    * makedumpfile.c
>    */
> diff --git a/diskdump.c b/diskdump.c
> index 2c284ff..d6c114d 100644
> --- a/diskdump.c
> +++ b/diskdump.c
> @@ -2745,6 +2745,8 @@ diskdump_device_dump_info(FILE *ofp)
>      }
>   }
> 
> +struct zram_pageflags zram_pageflags = { 0 };
> +
>   static void
>   zram_init(void)
>   {
> @@ -2754,6 +2756,17 @@ zram_init(void)
>      if (INVALID_MEMBER(zram_table_flag))
>          MEMBER_OFFSET_INIT(zram_table_flag, "zram_table_entry", "value");
>      STRUCT_SIZE_INIT(zram_table_entry, "zram_table_entry");
> +   MEMBER_OFFSET_INIT(zspoll_size_class, "zs_pool", "size_class");
> +   MEMBER_OFFSET_INIT(size_class_size, "size_class", "size");
> +   MEMBER_OFFSET_INIT(zspage_huge, "zspage", "huge");
> +
> +   if (!ZRAM_PAGEFLAG_VALUE(ZRAM_LOCK)) {
> +       ZRAM_PAGEFLAG_INIT(ZRAM_LOCK, 24);
> +       if (!enumerator_value("ZRAM_LOCK", &zram_pageflags.ZRAM_LOCK)
> +               && THIS_KERNEL_VERSION >= LINUX(6,1,0))
> +           ZRAM_PAGEFLAG_INIT(ZRAM_LOCK, PAGESHIFT() + 1);
> +       ZRAM_PAGEFLAG_INIT(ZRAM_SAME, ZRAM_PAGEFLAG_VALUE(ZRAM_LOCK) + 1);
> +   }
>   }
> 
>   static unsigned char *
> @@ -2761,9 +2774,11 @@ zram_object_addr(ulong pool, ulong handle, unsigned char *zram_buf)
>   {
>      ulong obj, off, class, page, zspage;
>      struct zspage zspage_s;
> +   struct zspage_5_17 zspage_5_17_s;
>      physaddr_t paddr;
>      unsigned int obj_idx, class_idx, size;
>      ulong pages[2], sizes[2];
> +   ulong zs_magic;
> 
>      readmem(handle, KVADDR, &obj, sizeof(void *), "zram entry", FAULT_ON_ERROR);
>      obj >>= OBJ_TAG_BITS;
> @@ -2772,11 +2787,19 @@ zram_object_addr(ulong pool, ulong handle, unsigned char *zram_buf)
> 
>      readmem(page + OFFSET(page_private), KVADDR, &zspage,
>              sizeof(void *), "page_private", FAULT_ON_ERROR);
> -   readmem(zspage, KVADDR, &zspage_s, sizeof(struct zspage), "zspage", FAULT_ON_ERROR);
> 
> -   class_idx = zspage_s.class;
> -   if (zspage_s.magic != ZSPAGE_MAGIC)
> -       error(FATAL, "zspage magic incorrect: %x\n", zspage_s.magic);
> +   if (VALID_MEMBER(zspage_huge)) {
> +       readmem(zspage, KVADDR, &zspage_5_17_s, sizeof(struct zspage_5_17), "zspage_5_17", FAULT_ON_ERROR);
> +       class_idx = zspage_5_17_s.class;
> +       zs_magic = zspage_5_17_s.magic;
> +   } else {
> +       readmem(zspage, KVADDR, &zspage_s, sizeof(struct zspage), "zspage", FAULT_ON_ERROR);
> +       class_idx = zspage_s.class;
> +       zs_magic = zspage_s.magic;
> +   }
> +
> +   if (zs_magic != ZSPAGE_MAGIC)
> +       error(FATAL, "zspage magic incorrect: %x\n", zs_magic);
> 
>      class = pool + OFFSET(zspoll_size_class);
>      class += (class_idx * sizeof(void *));
> @@ -2794,8 +2817,13 @@ zram_object_addr(ulong pool, ulong handle, unsigned char *zram_buf)
>      }
> 
>      pages[0] = page;
> -   readmem(page + OFFSET(page_freelist), KVADDR, &pages[1],
> -           sizeof(void *), "page_freelist", FAULT_ON_ERROR);
> +   if (VALID_MEMBER(page_freelist)) {
> +       readmem(page + OFFSET(page_freelist), KVADDR, &pages[1],
> +               sizeof(void *), "page_freelist", FAULT_ON_ERROR);
> +   } else {
> +       readmem(page + OFFSET(page_index), KVADDR, &pages[1],
> +               sizeof(void *), "page_index", FAULT_ON_ERROR);
> +   }
>      sizes[0] = PAGESIZE() - off;
>      sizes[1] = size - sizes[0];
>      if (!is_page_ptr(pages[0], &paddr)) {
> @@ -2812,9 +2840,15 @@ zram_object_addr(ulong pool, ulong handle, unsigned char *zram_buf)
>      readmem(paddr, PHYSADDR, zram_buf + sizes[0], sizes[1], "zram buffer[1]", FAULT_ON_ERROR);
> 
>   out:
> -   readmem(page, KVADDR, &obj, sizeof(void *), "page flags", FAULT_ON_ERROR);
> -   if (!(obj & (1<<10))) { //PG_OwnerPriv1 flag
> -       return (zram_buf + ZS_HANDLE_SIZE);
> +   if (!VALID_MEMBER(zspage_huge)) {
> +       readmem(page, KVADDR, &obj, sizeof(void *), "page flags", FAULT_ON_ERROR);
> +       if (!(obj & (1<<10))) { //PG_OwnerPriv1 flag
> +           return (zram_buf + ZS_HANDLE_SIZE);
> +       }
> +   } else {
> +       if (!zspage_5_17_s.huge) {
> +           return (zram_buf + ZS_HANDLE_SIZE);
> +       }
>      }
> 
>      return zram_buf;
> @@ -2929,6 +2963,7 @@ try_zram_decompress(ulonglong pte_val, unsigned char *buf, ulong len, ulonglong
>      unsigned char *outbuf = NULL;
>      ulong zram, zram_table_entry, sector, index, entry, flags, size,
>          outsize, off;
> +   unsigned long *same_buf = NULL;
> 
>      if (INVALID_MEMBER(zram_compressor)) {
>          zram_init();
> @@ -2992,11 +3027,18 @@ try_zram_decompress(ulonglong pte_val, unsigned char *buf, ulong len, ulonglong
>          sizeof(void *), "entry of table", FAULT_ON_ERROR);
>      readmem(zram_table_entry + OFFSET(zram_table_flag), KVADDR, &flags,
>          sizeof(void *), "zram_table_flag", FAULT_ON_ERROR);
> -   if (!entry || (flags & ZRAM_FLAG_SAME_BIT)) {
> -       memset(buf, entry, len);
> +
> +   if (!entry || flags & (1 << ZRAM_PAGEFLAG_VALUE(ZRAM_SAME))) {
> +       same_buf = (unsigned long *)GETBUF(PAGESIZE());
> +       for (int count = 0; count < PAGESIZE() / sizeof(unsigned long); count++) {
> +           same_buf[count] = entry;
> +       }
> +       memcpy(buf, same_buf + off, len);
> +       FREEBUF(same_buf);
>          goto out;
>      }
> -   size = flags & (ZRAM_FLAG_SHIFT -1);
> +
> +   size = flags & ((1 << ZRAM_PAGEFLAG_VALUE(ZRAM_LOCK)) - 1);
>      if (size == 0) {
>          len = 0;
>          goto out;
> --
> 2.39.0
> 
> Thanks.
> Guanyou.
> 
> --
> Crash-utility mailing list
> Crash-utility@xxxxxxxxxx
> https://listman.redhat.com/mailman/listinfo/crash-utility
> Contribution Guidelines: https://github.com/crash-utility/crash/wiki
> 
From 93cce82700015ed8b02a4ab55dcf8563e50b13e2 Mon Sep 17 00:00:00 2001
From: chenguanyou <chenguanyou@xxxxxxxxxx>
Date: Mon, 11 Sep 2023 20:59:39 +0800
Subject: [PATCH] Fix "rd" command to display data on zram on Linux 5.17 and later

Fix "rd" command to display data on zram on Linux 5.17 and later kernels
that contain commits
  a41ec880aa7b ("zsmalloc: move huge compressed obj from page to zspage"),
  ffedd09fa9b0 ("zsmalloc: Stop using slab fields in struct page"),
and on Linux 6.1 and later that contain commit
  f635725c3905 ("zram: do not waste zram_table_entry flags bits").

Also, fix a bug that sets the same "byte" by memset() instead to pages
containing the same "unsigned long" elements.

Before:
crash> mod -s zram zram.ko
     MODULE       NAME                                          BASE           SIZE  OBJECT FILE
ffffffde224db800  zram                                    ffffffde224d2000    57344  zram.ko
crash> mod -s zsmalloc zsmalloc.ko
     MODULE       NAME                                          BASE           SIZE  OBJECT FILE
ffffffde224c5180  zsmalloc                                ffffffde224bf000    40960  zsmalloc.ko
crash> rd 0x13d89fb0
rd: zspage magic incorrect: b0

After:
crash> rd 0x13d89fb0
        13d89fb0:  c2b54f7170883b20                     ;.pqO..

Signed-off-by: chenguanyou <chenguanyou@xxxxxxxxxx>
Signed-off-by: Kazuhito Hagio <k-hagio-ab@xxxxxxx>
---
 defs.h     | 16 ++++++++++++--
 diskdump.c | 61 +++++++++++++++++++++++++++++++++++++++++++++---------
 2 files changed, 65 insertions(+), 12 deletions(-)

diff --git a/defs.h b/defs.h
index 96a7a2a31471..122d3a969bf9 100644
--- a/defs.h
+++ b/defs.h
@@ -2225,6 +2225,7 @@ struct offset_table {                    /* stash of commonly-used offsets */
 	long module_memory_base;
 	long module_memory_size;
 	long irq_data_irq;
+	long zspage_huge;
 };
 
 struct size_table {         /* stash of commonly-used sizes */
@@ -7208,8 +7209,7 @@ ulong try_zram_decompress(ulonglong pte_val, unsigned char *buf, ulong len, ulon
 #define SECTOR_SHIFT     9
 #define SECTORS_PER_PAGE_SHIFT  (PAGESHIFT() - SECTOR_SHIFT)
 #define SECTORS_PER_PAGE        (1 << SECTORS_PER_PAGE_SHIFT)
-#define ZRAM_FLAG_SHIFT         (1<<24)
-#define ZRAM_FLAG_SAME_BIT      (1<<25)
+
 struct zspage {
     struct {
         unsigned int fullness : 2;
@@ -7221,6 +7221,18 @@ struct zspage {
     unsigned int freeobj;
 };
 
+struct zspage_5_17 {
+    struct {
+        unsigned int huge : 1;
+        unsigned int fullness : 2;
+        unsigned int class : 9;
+        unsigned int isolated : 3;
+        unsigned int magic : 8;
+    };
+    unsigned int inuse;
+    unsigned int freeobj;
+};
+
 /*
  * makedumpfile.c
  */
diff --git a/diskdump.c b/diskdump.c
index 2c284ff3f97f..dd7bd9f66fb5 100644
--- a/diskdump.c
+++ b/diskdump.c
@@ -2745,15 +2745,32 @@ diskdump_device_dump_info(FILE *ofp)
 	}
 }
 
+static long ZRAM_FLAG_SHIFT;
+static long ZRAM_FLAG_SAME_BIT;
+
 static void
 zram_init(void)
 {
+	long zram_lock;
+
 	MEMBER_OFFSET_INIT(zram_mempoll, "zram", "mem_pool");
 	MEMBER_OFFSET_INIT(zram_compressor, "zram", "compressor");
 	MEMBER_OFFSET_INIT(zram_table_flag, "zram_table_entry", "flags");
 	if (INVALID_MEMBER(zram_table_flag))
 		MEMBER_OFFSET_INIT(zram_table_flag, "zram_table_entry", "value");
 	STRUCT_SIZE_INIT(zram_table_entry, "zram_table_entry");
+	MEMBER_OFFSET_INIT(zspoll_size_class, "zs_pool", "size_class");
+	MEMBER_OFFSET_INIT(size_class_size, "size_class", "size");
+	MEMBER_OFFSET_INIT(zspage_huge, "zspage", "huge");
+
+	if (enumerator_value("ZRAM_LOCK", &zram_lock))
+		ZRAM_FLAG_SHIFT = 1 << zram_lock;
+	else if (THIS_KERNEL_VERSION >= LINUX(6,1,0))
+		ZRAM_FLAG_SHIFT = 1 << (PAGESHIFT() + 1);
+	else
+		ZRAM_FLAG_SHIFT = 1 << 24;
+
+	ZRAM_FLAG_SAME_BIT = ZRAM_FLAG_SHIFT << 1;
 }
 
 static unsigned char *
@@ -2761,9 +2778,11 @@ zram_object_addr(ulong pool, ulong handle, unsigned char *zram_buf)
 {
 	ulong obj, off, class, page, zspage;
 	struct zspage zspage_s;
+	struct zspage_5_17 zspage_5_17_s;
 	physaddr_t paddr;
 	unsigned int obj_idx, class_idx, size;
 	ulong pages[2], sizes[2];
+	ulong zs_magic;
 
 	readmem(handle, KVADDR, &obj, sizeof(void *), "zram entry", FAULT_ON_ERROR);
 	obj >>= OBJ_TAG_BITS;
@@ -2772,11 +2791,19 @@ zram_object_addr(ulong pool, ulong handle, unsigned char *zram_buf)
 
 	readmem(page + OFFSET(page_private), KVADDR, &zspage,
 			sizeof(void *), "page_private", FAULT_ON_ERROR);
-	readmem(zspage, KVADDR, &zspage_s, sizeof(struct zspage), "zspage", FAULT_ON_ERROR);
 
-	class_idx = zspage_s.class;
-	if (zspage_s.magic != ZSPAGE_MAGIC)
-		error(FATAL, "zspage magic incorrect: %x\n", zspage_s.magic);
+	if (VALID_MEMBER(zspage_huge)) {
+		readmem(zspage, KVADDR, &zspage_5_17_s, sizeof(struct zspage_5_17), "zspage_5_17", FAULT_ON_ERROR);
+		class_idx = zspage_5_17_s.class;
+		zs_magic = zspage_5_17_s.magic;
+	} else {
+		readmem(zspage, KVADDR, &zspage_s, sizeof(struct zspage), "zspage", FAULT_ON_ERROR);
+		class_idx = zspage_s.class;
+		zs_magic = zspage_s.magic;
+	}
+
+	if (zs_magic != ZSPAGE_MAGIC)
+		error(FATAL, "zspage magic incorrect: %x\n", zs_magic);
 
 	class = pool + OFFSET(zspoll_size_class);
 	class += (class_idx * sizeof(void *));
@@ -2794,8 +2821,13 @@ zram_object_addr(ulong pool, ulong handle, unsigned char *zram_buf)
 	}
 
 	pages[0] = page;
-	readmem(page + OFFSET(page_freelist), KVADDR, &pages[1],
-			sizeof(void *), "page_freelist", FAULT_ON_ERROR);
+	if (VALID_MEMBER(page_freelist))
+		readmem(page + OFFSET(page_freelist), KVADDR, &pages[1],
+				sizeof(void *), "page_freelist", FAULT_ON_ERROR);
+	else
+		readmem(page + OFFSET(page_index), KVADDR, &pages[1],
+				sizeof(void *), "page_index", FAULT_ON_ERROR);
+
 	sizes[0] = PAGESIZE() - off;
 	sizes[1] = size - sizes[0];
 	if (!is_page_ptr(pages[0], &paddr)) {
@@ -2812,9 +2844,13 @@ zram_object_addr(ulong pool, ulong handle, unsigned char *zram_buf)
 	readmem(paddr, PHYSADDR, zram_buf + sizes[0], sizes[1], "zram buffer[1]", FAULT_ON_ERROR);
 
 out:
-	readmem(page, KVADDR, &obj, sizeof(void *), "page flags", FAULT_ON_ERROR);
-	if (!(obj & (1<<10))) { //PG_OwnerPriv1 flag
-		return (zram_buf + ZS_HANDLE_SIZE);
+	if (VALID_MEMBER(zspage_huge)) {
+		if (!zspage_5_17_s.huge)
+			return (zram_buf + ZS_HANDLE_SIZE);
+	} else {
+		readmem(page, KVADDR, &obj, sizeof(void *), "page flags", FAULT_ON_ERROR);
+		if (!(obj & (1<<10))) // PG_OwnerPriv1 flag
+			return (zram_buf + ZS_HANDLE_SIZE);
 	}
 
 	return zram_buf;
@@ -2993,7 +3029,12 @@ try_zram_decompress(ulonglong pte_val, unsigned char *buf, ulong len, ulonglong
 	readmem(zram_table_entry + OFFSET(zram_table_flag), KVADDR, &flags,
 		sizeof(void *), "zram_table_flag", FAULT_ON_ERROR);
 	if (!entry || (flags & ZRAM_FLAG_SAME_BIT)) {
-		memset(buf, entry, len);
+		ulong *same_buf = (ulong *)GETBUF(PAGESIZE());
+		for (int count = 0; count < PAGESIZE() / sizeof(ulong); count++) {
+			same_buf[count] = entry;
+		}
+		memcpy(buf, same_buf + off, len);
+		FREEBUF(same_buf);
 		goto out;
 	}
 	size = flags & (ZRAM_FLAG_SHIFT -1);
-- 
2.31.1

--
Crash-utility mailing list
Crash-utility@xxxxxxxxxx
https://listman.redhat.com/mailman/listinfo/crash-utility
Contribution Guidelines: https://github.com/crash-utility/crash/wiki

[Index of Archives]     [Fedora Development]     [Fedora Desktop]     [Fedora SELinux]     [Yosemite News]     [KDE Users]     [Fedora Tools]

 

Powered by Linux