Re: [RFC v11][PATCH 05/13] Dump memory address space

[Mike Waychison <mikew@google.com>]

Comments below.

Oren Laadan wrote:
> For each VMA, there is a 'struct cr_vma'; if the VMA is file-mapped,
> it will be followed by the file name. Then comes the actual contents,
> in one or more chunk: each chunk begins with a header that specifies
> how many pages it holds, then the virtual addresses of all the dumped
> pages in that chunk, followed by the actual contents of all dumped
> pages. A header with zero number of pages marks the end of the contents.
> Then comes the next VMA and so on.
> 
> Changelog[v11]:
>   - Copy contents of 'init->fs->root' instead of pointing to them.
>   - Add missing test for VM_MAYSHARE when dumping memory
> 
> Changelog[v10]:
>   - Acquire dcache_lock around call to __d_path() in cr_fill_name()
> 
> Changelog[v9]:
>   - Introduce cr_ctx_checkpoint() for checkpoint-specific ctx setup
>   - Test if __d_path() changes mnt/dentry (when crossing filesystem
>     namespace boundary). for now cr_fill_fname() fails the checkpoint.
> 
> Changelog[v7]:
>   - Fix argument given to kunmap_atomic() in memory dump/restore
> 
> Changelog[v6]:
>   - Balance all calls to cr_hbuf_get() with matching cr_hbuf_put()
>     (even though it's not really needed)
> 
> Changelog[v5]:
>   - Improve memory dump code (following Dave Hansen's comments)
>   - Change dump format (and code) to allow chunks of <vaddrs, pages>
>     instead of one long list of each
>   - Fix use of follow_page() to avoid faulting in non-present pages
> 
> Changelog[v4]:
>   - Use standard list_... for cr_pgarr
> 
> Signed-off-by: Oren Laadan <orenl@cs.columbia.edu>
> Acked-by: Serge Hallyn <serue@us.ibm.com>
> Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
> ---
>  arch/x86/include/asm/checkpoint_hdr.h |    5 +
>  arch/x86/mm/checkpoint.c              |   31 ++
>  checkpoint/Makefile                   |    3 +-
>  checkpoint/checkpoint.c               |   88 ++++++
>  checkpoint/checkpoint_arch.h          |    2 +
>  checkpoint/checkpoint_mem.h           |   41 +++
>  checkpoint/ckpt_mem.c                 |  503 +++++++++++++++++++++++++++++++++
>  checkpoint/sys.c                      |   10 +
>  include/linux/checkpoint.h            |   12 +
>  include/linux/checkpoint_hdr.h        |   32 ++
>  10 files changed, 726 insertions(+), 1 deletions(-)
>  create mode 100644 checkpoint/checkpoint_mem.h
>  create mode 100644 checkpoint/ckpt_mem.c
> 
> diff --git a/arch/x86/include/asm/checkpoint_hdr.h b/arch/x86/include/asm/checkpoint_hdr.h
> index 6325062..33f4c70 100644
> --- a/arch/x86/include/asm/checkpoint_hdr.h
> +++ b/arch/x86/include/asm/checkpoint_hdr.h
> @@ -82,4 +82,9 @@ struct cr_hdr_cpu {
>  	/* thread_xstate contents follow (if used_math) */
>  } __attribute__((aligned(8)));
>  
> +struct cr_hdr_mm_context {
> +	__s16 ldt_entry_size;
> +	__s16 nldt;
> +} __attribute__((aligned(8)));
> +
>  #endif /* __ASM_X86_CKPT_HDR__H */
> diff --git a/arch/x86/mm/checkpoint.c b/arch/x86/mm/checkpoint.c
> index 8dd6d2d..757936e 100644
> --- a/arch/x86/mm/checkpoint.c
> +++ b/arch/x86/mm/checkpoint.c
> @@ -221,3 +221,34 @@ int cr_write_head_arch(struct cr_ctx *ctx)
>  
>  	return ret;
>  }
> +
> +/* dump the mm->context state */
> +int cr_write_mm_context(struct cr_ctx *ctx, struct mm_struct *mm, int parent)
> +{
> +	struct cr_hdr h;
> +	struct cr_hdr_mm_context *hh = cr_hbuf_get(ctx, sizeof(*hh));
> +	int ret;
> +
> +	h.type = CR_HDR_MM_CONTEXT;
> +	h.len = sizeof(*hh);
> +	h.parent = parent;
> +
> +	mutex_lock(&mm->context.lock);
> +
> +	hh->ldt_entry_size = LDT_ENTRY_SIZE;
> +	hh->nldt = mm->context.size;
> +
> +	cr_debug("nldt %d\n", hh->nldt);
> +
> +	ret = cr_write_obj(ctx, &h, hh);
> +	cr_hbuf_put(ctx, sizeof(*hh));
> +	if (ret < 0)
> +		goto out;
> +
> +	ret = cr_kwrite(ctx, mm->context.ldt,
> +			mm->context.size * LDT_ENTRY_SIZE);

Do we really want to emit anything under lock?  I realize that this 
patch goes and does a ton of writes with mmap_sem held for read -- is 
this ok?

> +
> + out:
> +	mutex_unlock(&mm->context.lock);
> +	return ret;
> +}
> diff --git a/checkpoint/Makefile b/checkpoint/Makefile
> index d2df68c..3a0df6d 100644
> --- a/checkpoint/Makefile
> +++ b/checkpoint/Makefile
> @@ -2,4 +2,5 @@
>  # Makefile for linux checkpoint/restart.
>  #
>  
> -obj-$(CONFIG_CHECKPOINT_RESTART) += sys.o checkpoint.o restart.o
> +obj-$(CONFIG_CHECKPOINT_RESTART) += sys.o checkpoint.o restart.o \
> +		ckpt_mem.o
> diff --git a/checkpoint/checkpoint.c b/checkpoint/checkpoint.c
> index 17cc8d2..56d0ec2 100644
> --- a/checkpoint/checkpoint.c
> +++ b/checkpoint/checkpoint.c
> @@ -13,6 +13,7 @@
>  #include <linux/time.h>
>  #include <linux/fs.h>
>  #include <linux/file.h>
> +#include <linux/fdtable.h>
>  #include <linux/dcache.h>
>  #include <linux/mount.h>
>  #include <linux/utsname.h>
> @@ -75,6 +76,66 @@ int cr_write_string(struct cr_ctx *ctx, char *str, int len)
>  	return cr_write_obj(ctx, &h, str);
>  }
>  
> +/**
> + * cr_fill_fname - return pathname of a given file
> + * @path: path name
> + * @root: relative root
> + * @buf: buffer for pathname
> + * @n: buffer length (in) and pathname length (out)
> + */
> +static char *
> +cr_fill_fname(struct path *path, struct path *root, char *buf, int *n)
> +{
> +	struct path tmp = *root;
> +	char *fname;
> +
> +	BUG_ON(!buf);
> +	spin_lock(&dcache_lock);
> +	fname = __d_path(path, &tmp, buf, *n);
> +	spin_unlock(&dcache_lock);
> +	if (!IS_ERR(fname))
> +		*n = (buf + (*n) - fname);
> +	/*
> +	 * FIXME: if __d_path() changed these, it must have stepped out of
> +	 * init's namespace. Since currently we require a unified namespace
> +	 * within the container: simply fail.
> +	 */
> +	if (tmp.mnt != root->mnt || tmp.dentry != root->dentry)
> +		fname = ERR_PTR(-EBADF);
> +
> +	return fname;
> +}
> +
> +/**
> + * cr_write_fname - write a file name
> + * @ctx: checkpoint context
> + * @path: path name
> + * @root: relative root
> + */
> +int cr_write_fname(struct cr_ctx *ctx, struct path *path, struct path *root)
> +{
> +	struct cr_hdr h;
> +	char *buf, *fname;
> +	int ret, flen;
> +
> +	flen = PATH_MAX;
> +	buf = kmalloc(flen, GFP_KERNEL);
> +	if (!buf)
> +		return -ENOMEM;
> +
> +	fname = cr_fill_fname(path, root, buf, &flen);
> +	if (!IS_ERR(fname)) {
> +		h.type = CR_HDR_FNAME;
> +		h.len = flen;
> +		h.parent = 0;
> +		ret = cr_write_obj(ctx, &h, fname);
> +	} else
> +		ret = PTR_ERR(fname);
> +
> +	kfree(buf);
> +	return ret;
> +}
> +
>  /* write the checkpoint header */
>  static int cr_write_head(struct cr_ctx *ctx)
>  {
> @@ -168,6 +229,10 @@ static int cr_write_task(struct cr_ctx *ctx, struct task_struct *t)
>  	cr_debug("task_struct: ret %d\n", ret);
>  	if (ret < 0)
>  		goto out;
> +	ret = cr_write_mm(ctx, t);
> +	cr_debug("memory: ret %d\n", ret);
> +	if (ret < 0)
> +		goto out;
>  	ret = cr_write_thread(ctx, t);
>  	cr_debug("thread: ret %d\n", ret);
>  	if (ret < 0)
> @@ -178,10 +243,33 @@ static int cr_write_task(struct cr_ctx *ctx, struct task_struct *t)
>  	return ret;
>  }
>  
> +static int cr_ctx_checkpoint(struct cr_ctx *ctx, pid_t pid)
> +{
> +	struct fs_struct *fs;
> +
> +	ctx->root_pid = pid;
> +
> +	/*
> +	 * assume checkpointer is in container's root vfs
> +	 * FIXME: this works for now, but will change with real containers
> +	 */
> +
> +	fs = current->fs;
> +	read_lock(&fs->lock);
> +	ctx->fs_mnt = fs->root;
> +	path_get(&ctx->fs_mnt);
> +	read_unlock(&fs->lock);
> +
> +	return 0;

Spurious return value?

> +}
> +
>  int do_checkpoint(struct cr_ctx *ctx, pid_t pid)
>  {
>  	int ret;
>  
> +	ret = cr_ctx_checkpoint(ctx, pid);
> +	if (ret < 0)
> +		goto out;
>  	ret = cr_write_head(ctx);
>  	if (ret < 0)
>  		goto out;
> diff --git a/checkpoint/checkpoint_arch.h b/checkpoint/checkpoint_arch.h
> index ada1369..f06c7eb 100644
> --- a/checkpoint/checkpoint_arch.h
> +++ b/checkpoint/checkpoint_arch.h
> @@ -3,6 +3,8 @@
>  extern int cr_write_head_arch(struct cr_ctx *ctx);
>  extern int cr_write_thread(struct cr_ctx *ctx, struct task_struct *t);
>  extern int cr_write_cpu(struct cr_ctx *ctx, struct task_struct *t);
> +extern int cr_write_mm_context(struct cr_ctx *ctx,
> +			       struct mm_struct *mm, int parent);
>  
>  extern int cr_read_head_arch(struct cr_ctx *ctx);
>  extern int cr_read_thread(struct cr_ctx *ctx);
> diff --git a/checkpoint/checkpoint_mem.h b/checkpoint/checkpoint_mem.h
> new file mode 100644
> index 0000000..85546f4
> --- /dev/null
> +++ b/checkpoint/checkpoint_mem.h
> @@ -0,0 +1,41 @@
> +#ifndef _CHECKPOINT_CKPT_MEM_H_
> +#define _CHECKPOINT_CKPT_MEM_H_
> +/*
> + *  Generic container checkpoint-restart
> + *
> + *  Copyright (C) 2008 Oren Laadan
> + *
> + *  This file is subject to the terms and conditions of the GNU General Public
> + *  License.  See the file COPYING in the main directory of the Linux
> + *  distribution for more details.
> + */
> +
> +#include <linux/mm_types.h>
> +
> +/*
> + * page-array chains: each cr_pgarr describes a set of <strcut page *,vaddr>

struct

> + * tuples (where vaddr is the virtual address of a page in a particular mm).
> + * Specifically, we use separate arrays so that all vaddrs can be written
> + * and read at once.
> + */
> +
> +struct cr_pgarr {
> +	unsigned long *vaddrs;
> +	struct page **pages;
> +	unsigned int nr_used;
> +	struct list_head list;
> +};
> +
> +#define CR_PGARR_TOTAL  (PAGE_SIZE / sizeof(void *))
> +#define CR_PGARR_CHUNK  (4 * CR_PGARR_TOTAL)
> +
> +extern void cr_pgarr_free(struct cr_ctx *ctx);
> +extern struct cr_pgarr *cr_pgarr_current(struct cr_ctx *ctx);
> +extern void cr_pgarr_reset_all(struct cr_ctx *ctx);
> +
> +static inline int cr_pgarr_is_full(struct cr_pgarr *pgarr)
> +{
> +	return (pgarr->nr_used == CR_PGARR_TOTAL);
> +}
> +
> +#endif /* _CHECKPOINT_CKPT_MEM_H_ */
> diff --git a/checkpoint/ckpt_mem.c b/checkpoint/ckpt_mem.c
> new file mode 100644
> index 0000000..a2fcdbf
> --- /dev/null
> +++ b/checkpoint/ckpt_mem.c
> @@ -0,0 +1,503 @@
> +/*
> + *  Checkpoint memory contents
> + *
> + *  Copyright (C) 2008 Oren Laadan
> + *
> + *  This file is subject to the terms and conditions of the GNU General Public
> + *  License.  See the file COPYING in the main directory of the Linux
> + *  distribution for more details.
> + */
> +
> +#include <linux/kernel.h>
> +#include <linux/sched.h>
> +#include <linux/slab.h>
> +#include <linux/file.h>
> +#include <linux/pagemap.h>
> +#include <linux/mm_types.h>
> +#include <linux/checkpoint.h>
> +#include <linux/checkpoint_hdr.h>
> +
> +#include "checkpoint_arch.h"
> +#include "checkpoint_mem.h"
> +
> +/*
> + * utilities to alloc, free, and handle 'struct cr_pgarr' (page-arrays)
> + * (common to ckpt_mem.c and rstr_mem.c).
> + *
> + * The checkpoint context structure has two members for page-arrays:
> + *   ctx->pgarr_list: list head of the page-array chain

What's the second member?

> + *
> + * During checkpoint (and restart) the chain tracks the dirty pages (page
> + * pointer and virtual address) of each MM. For a particular MM, these are
> + * always added to the head of the page-array chain (ctx->pgarr_list).
> + * This "current" page-array advances as necessary, and new page-array
> + * descriptors are allocated on-demand. Before the next chunk of pages,
> + * the chain is reset but not freed (that is, dereference page pointers).
> + */
> +
> +/* return first page-array in the chain */
> +static inline struct cr_pgarr *cr_pgarr_first(struct cr_ctx *ctx)
> +{
> +	if (list_empty(&ctx->pgarr_list))
> +		return NULL;
> +	return list_first_entry(&ctx->pgarr_list, struct cr_pgarr, list);
> +}
> +
> +/* release pages referenced by a page-array */
> +static void cr_pgarr_release_pages(struct cr_pgarr *pgarr)
> +{
> +	int i;
> +
> +	cr_debug("nr_used %d\n", pgarr->nr_used);
> +	/*
> +	 * although both checkpoint and restart use 'nr_used', we only
> +	 * collect pages during checkpoint; in restart we simply return
> +	 */
> +	if (!pgarr->pages)
> +		return;
> +	for (i = pgarr->nr_used; i--; /**/)
> +		page_cache_release(pgarr->pages[i]);

This is sorta hard to read (and non-intuitive).  Is it easier to do: 

 

for (i = 0; i < pgarr->nr_used; i++) 

	page_cache_release(pgarr->pages[i]);
 

It shouldn't matter what order you release the pages in..

> +}
> +
> +/* free a single page-array object */
> +static void cr_pgarr_free_one(struct cr_pgarr *pgarr)
> +{
> +	cr_pgarr_release_pages(pgarr);
> +	kfree(pgarr->pages);
> +	kfree(pgarr->vaddrs);
> +	kfree(pgarr);
> +}
> +
> +/* free a chain of page-arrays */
> +void cr_pgarr_free(struct cr_ctx *ctx)
> +{
> +	struct cr_pgarr *pgarr, *tmp;
> +
> +	list_for_each_entry_safe(pgarr, tmp, &ctx->pgarr_list, list) {
> +		list_del(&pgarr->list);
> +		cr_pgarr_free_one(pgarr);
> +	}
> +}
> +
> +/* allocate a single page-array object */
> +static struct cr_pgarr *cr_pgarr_alloc_one(unsigned long flags)
> +{
> +	struct cr_pgarr *pgarr;
> +
> +	pgarr = kzalloc(sizeof(*pgarr), GFP_KERNEL);
> +	if (!pgarr)
> +		return NULL;
> +
> +	pgarr->vaddrs = kmalloc(CR_PGARR_TOTAL * sizeof(unsigned long),
You used PAGE_SIZE / sizeof(void *) above.   Why not __get_free_page()?

> +				GFP_KERNEL);
> +	if (!pgarr->vaddrs)
> +		goto nomem;
> +
> +	/* pgarr->pages is needed only for checkpoint */
> +	if (flags & CR_CTX_CKPT) {
> +		pgarr->pages = kmalloc(CR_PGARR_TOTAL * sizeof(struct page *),
> +				       GFP_KERNEL);
> +		if (!pgarr->pages)
> +			goto nomem;
> +	}
> +
> +	return pgarr;
> +
> + nomem:
> +	cr_pgarr_free_one(pgarr);
> +	return NULL;
> +}
> +
> +/* cr_pgarr_current - return the next available page-array in the chain
> + * @ctx: checkpoint context
> + *
> + * Returns the first page-array in the list that has space. Extends the
> + * list if none has space.
> + */
> +struct cr_pgarr *cr_pgarr_current(struct cr_ctx *ctx)
> +{
> +	struct cr_pgarr *pgarr;
> +
> +	pgarr = cr_pgarr_first(ctx);
> +	if (pgarr && !cr_pgarr_is_full(pgarr))
> +		goto out;
> +	pgarr = cr_pgarr_alloc_one(ctx->flags);
> +	if (!pgarr)
> +		goto out;
> +	list_add(&pgarr->list, &ctx->pgarr_list);
> + out:
> +	return pgarr;
> +}
> +
> +/* reset the page-array chain (dropping page references if necessary) */
> +void cr_pgarr_reset_all(struct cr_ctx *ctx)
> +{
> +	struct cr_pgarr *pgarr;
> +
> +	list_for_each_entry(pgarr, &ctx->pgarr_list, list) {
> +		cr_pgarr_release_pages(pgarr);
> +		pgarr->nr_used = 0;
> +	}

This doesn't look right.  cr_pgarr_current only ever looks at the head 
of the list, so resetting a list with > 1 pgarr on it will mean the 
non-head elements in the list will go to waste.

> +}
> +
> +/*
> + * Checkpoint is outside the context of the checkpointee, so one cannot
> + * simply read pages from user-space. Instead, we scan the address space
> + * of the target to cherry-pick pages of interest. Selected pages are
> + * enlisted in a page-array chain (attached to the checkpoint context).
> + * To save their contents, each page is mapped to kernel memory and then
> + * dumped to the file descriptor.
> + */
> +
> +
> +/**
> + * cr_private_follow_page - return page pointer for dirty pages
> + * @vma - target vma
> + * @addr - page address
> + *
> + * Looks up the page that correspond to the address in the vma, and
> + * returns the page if it was modified (and grabs a reference to it),
> + * or otherwise returns NULL (or error).
> + *
> + * This function should _only_ called for private vma's.
> + */
> +static struct page *
> +cr_private_follow_page(struct vm_area_struct *vma, unsigned long addr)

s/cr_private_follow_page/cr_follow_page_private/ ?

Maybe even cr_dump_private_page?  The fact that it's following the page 
  tables down to the page is an implementation artifact and isn't really 
relevant to the semantics you want to express.

> +{
> +	struct page *page;
> +
> +	BUG_ON(vma->vm_flags & (VM_SHARED | VM_MAYSHARE));
> +

This BUG_ON shouldn't be needed if it's already done in 
cr_private_vma_fill_pgarr.

> +	/*
> +	 * simplified version of get_user_pages(): already have vma,
> +	 * only need FOLL_ANON, and (for now) ignore fault stats.
> +	 *
> +	 * follow_page() will return NULL if the page is not present
> +	 * (swapped), ZERO_PAGE(0) if the pte wasn't allocated, and
> +	 * the actual page pointer otherwise.
> +	 *
> +	 * FIXME: consolidate with get_user_pages()
> +	 */
> +
> +	cond_resched();
> +	while (!(page = follow_page(vma, addr, FOLL_ANON | FOLL_GET))) {
> +		int ret;
> +
> +		/* the page is swapped out - bring it in (optimize ?) */
> +		ret = handle_mm_fault(vma->vm_mm, vma, addr, 0);
> +		if (ret & VM_FAULT_ERROR) {
> +			if (ret & VM_FAULT_OOM)
> +				return ERR_PTR(-ENOMEM);
> +			else if (ret & VM_FAULT_SIGBUS)
> +				return ERR_PTR(-EFAULT);
> +			else
> +				BUG();
> +			break;
> +		}
> +		cond_resched();
> +	}
> +
> +	if (IS_ERR(page))
> +		return page;
> +
> +	/*
> +	 * We only care about dirty pages: either non-zero page, or
> +	 * file-backed (copy-on-write) that were touched. For the latter,
> +	 * the page_mapping() will be unset because it will no longer be
> +	 * mapped to the original file  after having been modified.
> +	 */
> +	if (page == ZERO_PAGE(0)) {
> +		/* this is the zero page: ignore */
> +		page_cache_release(page);
> +		page = NULL;
> +	} else if (vma->vm_file && (page_mapping(page) != NULL)) {
> +		/* file backed clean cow: ignore */

Probably better to describe 'why' it can be ignored here.


> +		page_cache_release(page);
> +		page = NULL;
> +	}
> +
> +	return page;
> +}
> +
> +/**
> + * cr_private_vma_fill_pgarr - fill a page-array with addr/page tuples
> + * @ctx - checkpoint context
> + * @pgarr - page-array to fill
> + * @vma - vma to scan
> + * @start - start address (updated)
> + *
> + * Returns the number of pages collected
> + */
> +static int
> +cr_private_vma_fill_pgarr(struct cr_ctx *ctx, struct cr_pgarr *pgarr,
> +			  struct vm_area_struct *vma, unsigned long *start)

This is sorta nasty because you shouldn't need to call into this routine 
with a container.  It should be able to enqueue the (vaddr, page) tuple 
directly on the ctx.  Doing so would also abstract out the pgarr 
management at this level and make the code a lot simpler.

> +{
> +	unsigned long end = vma->vm_end;
> +	unsigned long addr = *start;
> +	int orig_used = pgarr->nr_used;
> +
> +	/* this function is only for private memory (anon or file-mapped) */
> +	BUG_ON(vma->vm_flags & (VM_SHARED | VM_MAYSHARE));
> +
> +	while (addr < end) {
> +		struct page *page;
> +
> +		page = cr_private_follow_page(vma, addr);
> +		if (IS_ERR(page))
> +			return PTR_ERR(page);
> +
> +		if (page) {
> +			pgarr->pages[pgarr->nr_used] = page;
> +			pgarr->vaddrs[pgarr->nr_used] = addr;
> +			pgarr->nr_used++;

Should be something like:

ret = cr_ctx_append_page(ctx, addr, page);
if (ret < 0)
   goto out;

> +		}
> +
> +		addr += PAGE_SIZE;
> +
> +		if (cr_pgarr_is_full(pgarr))
> +			break;
> +	}
> +
> +	*start = addr;
> +	return pgarr->nr_used - orig_used;
> +}
> +
> +/* dump contents of a pages: use kmap_atomic() to avoid TLB flush */
> +static int cr_page_write(struct cr_ctx *ctx, struct page *page, char *buf)
> +{
> +	void *ptr;
> +
> +	ptr = kmap_atomic(page, KM_USER1);
> +	memcpy(buf, ptr, PAGE_SIZE);
> +	kunmap_atomic(ptr, KM_USER1);
> +
> +	return cr_kwrite(ctx, buf, PAGE_SIZE);
> +}
> +
> +/**
> + * cr_vma_dump_pages - dump pages listed in the ctx page-array chain
> + * @ctx - checkpoint context
> + * @total - total number of pages
> + *
> + * First dump all virtual addresses, followed by the contents of all pages
> + */
> +static int cr_vma_dump_pages(struct cr_ctx *ctx, int total)
> +{
> +	struct cr_pgarr *pgarr;
> +	char *buf;
> +	int i, ret = 0;
> +
> +	if (!total)
> +		return 0;
> +
> +	list_for_each_entry_reverse(pgarr, &ctx->pgarr_list, list) {
> +		ret = cr_kwrite(ctx, pgarr->vaddrs,
> +				pgarr->nr_used * sizeof(*pgarr->vaddrs));
> +		if (ret < 0)
> +			return ret;
> +	}
> +
> +	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);

__get_free_page()

> +	if (!buf)
> +		return -ENOMEM;
> +
> +	list_for_each_entry_reverse(pgarr, &ctx->pgarr_list, list) {
> +		for (i = 0; i < pgarr->nr_used; i++) {
> +			ret = cr_page_write(ctx, pgarr->pages[i], buf);
> +			if (ret < 0)
> +				goto out;
> +		}
> +	}
> +
> + out:
> +	kfree(buf);
> +	return ret;
> +}
> +
> +/**
> + * cr_write_private_vma_contents - dump contents of a VMA with private memory
> + * @ctx - checkpoint context
> + * @vma - vma to scan
> + *
> + * Collect lists of pages that needs to be dumped, and corresponding
> + * virtual addresses into ctx->pgarr_list page-array chain. Then dump
> + * the addresses, followed by the page contents.
> + */
> +static int
> +cr_write_private_vma_contents(struct cr_ctx *ctx, struct vm_area_struct *vma)
> +{
> +	struct cr_hdr h;
> +	struct cr_hdr_pgarr *hh;
> +	unsigned long addr = vma->vm_start;
> +	struct cr_pgarr *pgarr;
> +	unsigned long cnt = 0;
> +	int ret;
> +
> +	/*
> +	 * Work iteratively, collecting and dumping at most CR_PGARR_CHUNK
> +	 * in each round. Each iterations is divided into two steps:
> +	 *
> +	 * (1) scan: scan through the PTEs of the vma to collect the pages
> +	 * to dump (later we'll also make them COW), while keeping a list
> +	 * of pages and their corresponding addresses on ctx->pgarr_list.
> +	 *
> +	 * (2) dump: write out a header specifying how many pages, followed
> +	 * by the addresses of all pages in ctx->pgarr_list, followed by
> +	 * the actual contents of all pages. (Then, release the references
> +	 * to the pages and reset the page-array chain).
> +	 *
> +	 * (This split makes the logic simpler by first counting the pages
> +	 * that need saving. More importantly, it allows for a future
> +	 * optimization that will reduce application downtime by deferring
> +	 * the actual write-out of the data to after the application is
> +	 * allowed to resume execution).
> +	 *
> +	 * After dumpting the entire contents, conclude with a header that
> +	 * specifies 0 pages to mark the end of the contents.
> +	 */
> +
> +	h.type = CR_HDR_PGARR;
> +	h.len = sizeof(*hh);
> +	h.parent = 0;
> +
> +	while (addr < vma->vm_end) {
> +		pgarr = cr_pgarr_current(ctx);
> +		if (!pgarr)
> +			return -ENOMEM;
> +		ret = cr_private_vma_fill_pgarr(ctx, pgarr, vma, &addr);
> +		if (ret < 0)
> +			return ret;
> +		cnt += ret;
> +
> +		/* did we complete a chunk, or is this the last chunk ? */
> +		if (cnt >= CR_PGARR_CHUNK || (cnt && addr == vma->vm_end)) {
> +			hh = cr_hbuf_get(ctx, sizeof(*hh));
> +			hh->nr_pages = cnt;
> +			ret = cr_write_obj(ctx, &h, hh);
> +			cr_hbuf_put(ctx, sizeof(*hh));
> +			if (ret < 0)
> +				return ret;
> +
> +			ret = cr_vma_dump_pages(ctx, cnt);
> +			if (ret < 0)
> +				return ret;
> +
> +			cr_pgarr_reset_all(ctx);
> +		}
> +	}
> +
> +	/* mark end of contents with header saying "0" pages */
> +	hh = cr_hbuf_get(ctx, sizeof(*hh));
> +	hh->nr_pages = 0;
> +	ret = cr_write_obj(ctx, &h, hh);
> +	cr_hbuf_put(ctx, sizeof(*hh));
> +
> +	return ret;
> +}
> +
> +static int cr_write_vma(struct cr_ctx *ctx, struct vm_area_struct *vma)
> +{
> +	struct cr_hdr h;
> +	struct cr_hdr_vma *hh = cr_hbuf_get(ctx, sizeof(*hh));
> +	int vma_type, ret;
> +
> +	h.type = CR_HDR_VMA;
> +	h.len = sizeof(*hh);
> +	h.parent = 0;
> +
> +	hh->vm_start = vma->vm_start;
> +	hh->vm_end = vma->vm_end;
> +	hh->vm_page_prot = vma->vm_page_prot.pgprot;
> +	hh->vm_flags = vma->vm_flags;
> +	hh->vm_pgoff = vma->vm_pgoff;
> +
> +#define CR_BAD_VM_FLAGS  \
> +	(VM_SHARED | VM_MAYSHARE | VM_IO | VM_HUGETLB | VM_NONLINEAR)
> +
> +	if (vma->vm_flags & CR_BAD_VM_FLAGS) {
> +		pr_warning("c/r: unsupported VMA %#lx\n", vma->vm_flags);
> +		cr_hbuf_put(ctx, sizeof(*hh));
> +		return -ENOSYS;
> +	}
> +

The following code should be broken into it's own function?  Handling of 
other types of memory will follow and will clutter this guy up.

> +	/* by default assume anon memory */
> +	vma_type = CR_VMA_ANON;
> +
> +	/*
> +	 * if there is a backing file, assume private-mapped

Shouldn't need to assume anything as you checked for VM_MAYSHARE and 
VM_SHARED above.

> +	 * (FIXME: check if the file is unlinked)
> +	 */
> +	if (vma->vm_file)
> +		vma_type = CR_VMA_FILE;
> +
> +	hh->vma_type = vma_type;
> +
> +	ret = cr_write_obj(ctx, &h, hh);
> +	cr_hbuf_put(ctx, sizeof(*hh));
> +	if (ret < 0)
> +		return ret;
> +
> +	/* save the file name, if relevant */

s/, if relevant//

> +	if (vma->vm_file) {
> +		ret = cr_write_fname(ctx, &vma->vm_file->f_path, &ctx->fs_mnt);

Why is this using a filename, rather than a reference to a file? 
Shouldn't this use the logic in patch 8/13?

> +		if (ret < 0)
> +			return ret;
> +	}
> +
> +	return cr_write_private_vma_contents(ctx, vma);
> +}
> +
> +int cr_write_mm(struct cr_ctx *ctx, struct task_struct *t)
> +{
> +	struct cr_hdr h;
> +	struct cr_hdr_mm *hh = cr_hbuf_get(ctx, sizeof(*hh));
> +	struct mm_struct *mm;
> +	struct vm_area_struct *vma;
> +	int objref, ret;
> +
> +	h.type = CR_HDR_MM;
> +	h.len = sizeof(*hh);
> +	h.parent = task_pid_vnr(t);
> +
> +	mm = get_task_mm(t);
> +
> +	objref = 0;	/* will be meaningful with multiple processes */
> +	hh->objref = objref;
> +
> +	down_read(&mm->mmap_sem);
> +
> +	hh->start_code = mm->start_code;
> +	hh->end_code = mm->end_code;
> +	hh->start_data = mm->start_data;
> +	hh->end_data = mm->end_data;
> +	hh->start_brk = mm->start_brk;
> +	hh->brk = mm->brk;
> +	hh->start_stack = mm->start_stack;
> +	hh->arg_start = mm->arg_start;
> +	hh->arg_end = mm->arg_end;
> +	hh->env_start = mm->env_start;
> +	hh->env_end = mm->env_end;
> +
> +	hh->map_count = mm->map_count;
> +
> +	/* FIX: need also mm->flags */
> +
> +	ret = cr_write_obj(ctx, &h, hh);
> +	cr_hbuf_put(ctx, sizeof(*hh));
> +	if (ret < 0)
> +		goto out;
> +
> +	/* write the vma's */
> +	for (vma = mm->mmap; vma; vma = vma->vm_next) {
> +		ret = cr_write_vma(ctx, vma);
> +		if (ret < 0)
> +			goto out;
> +	}
> +
> +	ret = cr_write_mm_context(ctx, mm, objref);
> +
> + out:
> +	up_read(&mm->mmap_sem);
> +	mmput(mm);
> +	return ret;
> +}
> diff --git a/checkpoint/sys.c b/checkpoint/sys.c
> index bd14ef9..c547a1c 100644
> --- a/checkpoint/sys.c
> +++ b/checkpoint/sys.c
> @@ -16,6 +16,8 @@
>  #include <linux/capability.h>
>  #include <linux/checkpoint.h>
>  
> +#include "checkpoint_mem.h"
> +
>  /*
>   * Helpers to write(read) from(to) kernel space to(from) the checkpoint
>   * image file descriptor (similar to how a core-dump is performed).
> @@ -131,7 +133,13 @@ static void cr_ctx_free(struct cr_ctx *ctx)
>  {
>  	if (ctx->file)
>  		fput(ctx->file);
> +
>  	kfree(ctx->hbuf);
> +
> +	path_put(&ctx->fs_mnt);		/* safe with NULL pointers */
> +
> +	cr_pgarr_free(ctx);
> +
>  	kfree(ctx);
>  }
>  
> @@ -146,6 +154,8 @@ static struct cr_ctx *cr_ctx_alloc(int fd, unsigned long flags)
>  
>  	ctx->flags = flags;
>  
> +	INIT_LIST_HEAD(&ctx->pgarr_list);
> +
>  	err = -EBADF;
>  	ctx->file = fget(fd);
>  	if (!ctx->file)
> diff --git a/include/linux/checkpoint.h b/include/linux/checkpoint.h
> index 63f298f..4e97f9f 100644
> --- a/include/linux/checkpoint.h
> +++ b/include/linux/checkpoint.h
> @@ -10,6 +10,9 @@
>   *  distribution for more details.
>   */
>  
> +#include <linux/path.h>
> +#include <linux/fs.h>
> +
>  #define CR_VERSION  1
>  
>  struct cr_ctx {
> @@ -25,6 +28,10 @@ struct cr_ctx {
>  
>  	void *hbuf;		/* temporary buffer for headers */
>  	int hpos;		/* position in headers buffer */
> +
> +	struct list_head pgarr_list;	/* page array to dump VMA contents */
> +
> +	struct path fs_mnt;	/* container root (FIXME) */
>  };
>  
>  /* cr_ctx: flags */
> @@ -42,6 +49,8 @@ struct cr_hdr;
>  extern int cr_write_obj(struct cr_ctx *ctx, struct cr_hdr *h, void *buf);
>  extern int cr_write_buffer(struct cr_ctx *ctx, void *buf, int len);
>  extern int cr_write_string(struct cr_ctx *ctx, char *str, int len);
> +extern int cr_write_fname(struct cr_ctx *ctx,
> +			  struct path *path, struct path *root);
>  
>  extern int cr_read_obj(struct cr_ctx *ctx, struct cr_hdr *h, void *buf, int n);
>  extern int cr_read_obj_type(struct cr_ctx *ctx, void *buf, int len, int type);
> @@ -50,7 +59,10 @@ extern int cr_read_buffer(struct cr_ctx *ctx, void *buf, int *len);
>  extern int cr_read_string(struct cr_ctx *ctx, char *str, int len);
>  
>  extern int do_checkpoint(struct cr_ctx *ctx, pid_t pid);
> +extern int cr_write_mm(struct cr_ctx *ctx, struct task_struct *t);
> +
>  extern int do_restart(struct cr_ctx *ctx, pid_t pid);
> +extern int cr_read_mm(struct cr_ctx *ctx);
>  
>  #define cr_debug(fmt, args...)  \
>  	pr_debug("[%d:c/r:%s] " fmt, task_pid_vnr(current), __func__, ## args)
> diff --git a/include/linux/checkpoint_hdr.h b/include/linux/checkpoint_hdr.h
> index b74b5f9..d78f0f1 100644
> --- a/include/linux/checkpoint_hdr.h
> +++ b/include/linux/checkpoint_hdr.h
> @@ -34,6 +34,7 @@ enum {
>  	CR_HDR_HEAD_ARCH,
>  	CR_HDR_BUFFER,
>  	CR_HDR_STRING,
> +	CR_HDR_FNAME,
>  
>  	CR_HDR_TASK = 101,
>  	CR_HDR_THREAD,
> @@ -41,6 +42,7 @@ enum {
>  
>  	CR_HDR_MM = 201,
>  	CR_HDR_VMA,
> +	CR_HDR_PGARR,
>  	CR_HDR_MM_CONTEXT,
>  
>  	CR_HDR_TAIL = 5001
> @@ -75,4 +77,34 @@ struct cr_hdr_task {
>  	__s32 task_comm_len;
>  } __attribute__((aligned(8)));
>  
> +struct cr_hdr_mm {
> +	__u32 objref;		/* identifier for shared objects */
> +	__u32 map_count;
> +
> +	__u64 start_code, end_code, start_data, end_data;
> +	__u64 start_brk, brk, start_stack;
> +	__u64 arg_start, arg_end, env_start, env_end;
> +} __attribute__((aligned(8)));
> +
> +/* vma subtypes */
> +enum vm_type {
> +	CR_VMA_ANON = 1,
> +	CR_VMA_FILE

We need to figure out what MAP_SHARED | MAP_ANONYMOUS should be exposed 
as in this setup (much in the same way we need to start defining what 
shm mappings look like).  Internally, they are 'file-backed', but to 
userland, they aren't.
 

Thoughts?

> +};
> +
> +struct cr_hdr_vma {
> +	__u32 vma_type;
> +	__u32 _padding;

Why padding?

> +
> +	__u64 vm_start;
> +	__u64 vm_end;
> +	__u64 vm_page_prot;
> +	__u64 vm_flags;
> +	__u64 vm_pgoff;
> +} __attribute__((aligned(8)));
> +
> +struct cr_hdr_pgarr {
> +	__u64 nr_pages;		/* number of pages to saved */
> +} __attribute__((aligned(8)));
> +
>  #endif /* _CHECKPOINT_CKPT_HDR_H_ */

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>