x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

When using eBPF to probe CS CALL instructions within a function,
a crash can be triggered.

The eBPF tool probes offset 257 of the __hrtimer_run_queues()
function:

<__hrtimer_run_queues+249>:  nopl   0x0(%rax,%rax,1)
<__hrtimer_run_queues+254>:  mov    %r14,%rdi
<__hrtimer_run_queues+257>:  cs call <__x86_indirect_thunk_r12>
<__hrtimer_run_queues+263>:  mov    %eax,%r12d
<__hrtimer_run_queues+266>:  xchg   %ax,%ax
<__hrtimer_run_queues+268>:  mov    %r13,%rdi

Which triggers this crash:

  BUG: unable to handle page fault for address: 00000000000f41c9
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 0 P4D 0
  Oops: 0002 [#1] SMP NOPTI
  CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P
  RIP: 0010:__hrtimer_run_queues+0x106/0x230

Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is
at the 6th byte of the above CS CALL instruction. Since the CS CALL
instruction occupies 6 bytes, the exception occurred in the middle of that
call instruction.

The root cause is that when using eBPF tools to probe in the middle of a
function, a kprobe with INT3 is used as the underlying implementation.

During single-step emulation of the original CALL instruction,
int3_emulate_call() assumes that the probed CALL instruction is 5 bytes
long. However, the actual CS-prefixed CALL instruction occupies 6 bytes,
so it constructs an incorrect exception return address. When the CPU
returns from the kprobe handler, the next instruction to be executed is at
the address of the last byte of that CS CALL instruction. Coincidentally,
starting from that address, the CPU fetches and decodes a completely
different instruction, which ultimately triggers a kernel crash.

Fix the issue by using the actual instruction length obtained from
the instruction decoder when constructing the exception return
address, rather than relying on the hardcoded CALL_INSN_SIZE macro.

[ mingo: Refined the changelog ]

Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step")
Suggested-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 336 insertions, 0 deletions

diff --git a/fs/jffs2/file.c b/fs/jffs2/file.c
new file mode 100644
index 000000000..1e18d3a79
--- /dev/null
+++ b/fs/jffs2/file.c
@@ -0,0 +1,336 @@
+/*
+ * JFFS2 -- Journalling Flash File System, Version 2.
+ *
+ * Copyright © 2001-2007 Red Hat, Inc.
+ * Copyright © 2004-2010 David Woodhouse <dwmw2@infradead.org>
+ *
+ * Created by David Woodhouse <dwmw2@infradead.org>
+ *
+ * For licensing information, see the file 'LICENCE' in this directory.
+ *
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/kernel.h>
+#include <linux/fs.h>
+#include <linux/filelock.h>
+#include <linux/time.h>
+#include <linux/pagemap.h>
+#include <linux/highmem.h>
+#include <linux/crc32.h>
+#include <linux/jffs2.h>
+#include "nodelist.h"
+
+static int jffs2_write_end(const struct kiocb *iocb,
+			   struct address_space *mapping,
+			   loff_t pos, unsigned len, unsigned copied,
+			   struct folio *folio, void *fsdata);
+static int jffs2_write_begin(const struct kiocb *iocb,
+			     struct address_space *mapping,
+			     loff_t pos, unsigned len,
+			     struct folio **foliop, void **fsdata);
+static int jffs2_read_folio(struct file *filp, struct folio *folio);
+
+int jffs2_fsync(struct file *filp, loff_t start, loff_t end, int datasync)
+{
+	struct inode *inode = filp->f_mapping->host;
+	struct jffs2_sb_info *c = JFFS2_SB_INFO(inode->i_sb);
+	int ret;
+
+	ret = file_write_and_wait_range(filp, start, end);
+	if (ret)
+		return ret;
+
+	inode_lock(inode);
+	/* Trigger GC to flush any pending writes for this inode */
+	jffs2_flush_wbuf_gc(c, inode->i_ino);
+	inode_unlock(inode);
+
+	return 0;
+}
+
+const struct file_operations jffs2_file_operations =
+{
+	.llseek =	generic_file_llseek,
+	.open =		generic_file_open,
+ 	.read_iter =	generic_file_read_iter,
+ 	.write_iter =	generic_file_write_iter,
+	.unlocked_ioctl=jffs2_ioctl,
+	.mmap_prepare =	generic_file_readonly_mmap_prepare,
+	.fsync =	jffs2_fsync,
+	.splice_read =	filemap_splice_read,
+	.splice_write = iter_file_splice_write,
+	.setlease =	generic_setlease,
+};
+
+/* jffs2_file_inode_operations */
+
+const struct inode_operations jffs2_file_inode_operations =
+{
+	.get_inode_acl =	jffs2_get_acl,
+	.set_acl =	jffs2_set_acl,
+	.setattr =	jffs2_setattr,
+	.listxattr =	jffs2_listxattr,
+};
+
+const struct address_space_operations jffs2_file_address_operations =
+{
+	.read_folio =	jffs2_read_folio,
+	.write_begin =	jffs2_write_begin,
+	.write_end =	jffs2_write_end,
+};
+
+static int jffs2_do_readpage_nolock(struct inode *inode, struct folio *folio)
+{
+	struct jffs2_inode_info *f = JFFS2_INODE_INFO(inode);
+	struct jffs2_sb_info *c = JFFS2_SB_INFO(inode->i_sb);
+	unsigned char *kaddr;
+	int ret;
+
+	jffs2_dbg(2, "%s(): ino #%llu, page at offset 0x%lx\n",
+		  __func__, inode->i_ino, folio->index << PAGE_SHIFT);
+
+	BUG_ON(!folio_test_locked(folio));
+
+	kaddr = kmap_local_folio(folio, 0);
+	ret = jffs2_read_inode_range(c, f, kaddr, folio->index << PAGE_SHIFT,
+				     PAGE_SIZE);
+	kunmap_local(kaddr);
+
+	if (!ret)
+		folio_mark_uptodate(folio);
+
+	flush_dcache_folio(folio);
+
+	jffs2_dbg(2, "readpage finished\n");
+	return ret;
+}
+
+int __jffs2_read_folio(struct file *file, struct folio *folio)
+{
+	int ret = jffs2_do_readpage_nolock(folio->mapping->host, folio);
+	folio_unlock(folio);
+	return ret;
+}
+
+static int jffs2_read_folio(struct file *file, struct folio *folio)
+{
+	struct jffs2_inode_info *f = JFFS2_INODE_INFO(folio->mapping->host);
+	int ret;
+
+	mutex_lock(&f->sem);
+	ret = __jffs2_read_folio(file, folio);
+	mutex_unlock(&f->sem);
+	return ret;
+}
+
+static int jffs2_write_begin(const struct kiocb *iocb,
+			     struct address_space *mapping,
+			     loff_t pos, unsigned len,
+			     struct folio **foliop, void **fsdata)
+{
+	struct folio *folio;
+	struct inode *inode = mapping->host;
+	struct jffs2_inode_info *f = JFFS2_INODE_INFO(inode);
+	struct jffs2_sb_info *c = JFFS2_SB_INFO(inode->i_sb);
+	pgoff_t index = pos >> PAGE_SHIFT;
+	int ret = 0;
+
+	jffs2_dbg(1, "%s()\n", __func__);
+
+	if (pos > inode->i_size) {
+		/* Make new hole frag from old EOF to new position */
+		struct jffs2_raw_inode ri;
+		struct jffs2_full_dnode *fn;
+		uint32_t alloc_len;
+
+		jffs2_dbg(1, "Writing new hole frag 0x%x-0x%x between current EOF and new position\n",
+			  (unsigned int)inode->i_size, (uint32_t)pos);
+
+		ret = jffs2_reserve_space(c, sizeof(ri), &alloc_len,
+					  ALLOC_NORMAL, JFFS2_SUMMARY_INODE_SIZE);
+		if (ret)
+			goto out_err;
+
+		mutex_lock(&f->sem);
+		memset(&ri, 0, sizeof(ri));
+
+		ri.magic = cpu_to_je16(JFFS2_MAGIC_BITMASK);
+		ri.nodetype = cpu_to_je16(JFFS2_NODETYPE_INODE);
+		ri.totlen = cpu_to_je32(sizeof(ri));
+		ri.hdr_crc = cpu_to_je32(crc32(0, &ri, sizeof(struct jffs2_unknown_node)-4));
+
+		ri.ino = cpu_to_je32(f->inocache->ino);
+		ri.version = cpu_to_je32(++f->highest_version);
+		ri.mode = cpu_to_jemode(inode->i_mode);
+		ri.uid = cpu_to_je16(i_uid_read(inode));
+		ri.gid = cpu_to_je16(i_gid_read(inode));
+		ri.isize = cpu_to_je32((uint32_t)pos);
+		ri.atime = ri.ctime = ri.mtime = cpu_to_je32(JFFS2_NOW());
+		ri.offset = cpu_to_je32(inode->i_size);
+		ri.dsize = cpu_to_je32((uint32_t)pos - inode->i_size);
+		ri.csize = cpu_to_je32(0);
+		ri.compr = JFFS2_COMPR_ZERO;
+		ri.node_crc = cpu_to_je32(crc32(0, &ri, sizeof(ri)-8));
+		ri.data_crc = cpu_to_je32(0);
+
+		fn = jffs2_write_dnode(c, f, &ri, NULL, 0, ALLOC_NORMAL);
+
+		if (IS_ERR(fn)) {
+			ret = PTR_ERR(fn);
+			jffs2_complete_reservation(c);
+			mutex_unlock(&f->sem);
+			goto out_err;
+		}
+		ret = jffs2_add_full_dnode_to_inode(c, f, fn);
+		if (f->metadata) {
+			jffs2_mark_node_obsolete(c, f->metadata->raw);
+			jffs2_free_full_dnode(f->metadata);
+			f->metadata = NULL;
+		}
+		if (ret) {
+			jffs2_dbg(1, "Eep. add_full_dnode_to_inode() failed in write_begin, returned %d\n",
+				  ret);
+			jffs2_mark_node_obsolete(c, fn->raw);
+			jffs2_free_full_dnode(fn);
+			jffs2_complete_reservation(c);
+			mutex_unlock(&f->sem);
+			goto out_err;
+		}
+		jffs2_complete_reservation(c);
+		inode->i_size = pos;
+		mutex_unlock(&f->sem);
+	}
+
+	/*
+	 * While getting a page and reading data in, lock c->alloc_sem until
+	 * the page is Uptodate. Otherwise GC task may attempt to read the same
+	 * page in read_cache_page(), which causes a deadlock.
+	 */
+	mutex_lock(&c->alloc_sem);
+	folio = __filemap_get_folio(mapping, index, FGP_WRITEBEGIN,
+			mapping_gfp_mask(mapping));
+	if (IS_ERR(folio)) {
+		ret = PTR_ERR(folio);
+		goto release_sem;
+	}
+	*foliop = folio;
+
+	/*
+	 * Read in the folio if it wasn't already present. Cannot optimize away
+	 * the whole folio write case until jffs2_write_end can handle the
+	 * case of a short-copy.
+	 */
+	if (!folio_test_uptodate(folio)) {
+		mutex_lock(&f->sem);
+		ret = jffs2_do_readpage_nolock(inode, folio);
+		mutex_unlock(&f->sem);
+		if (ret) {
+			folio_unlock(folio);
+			folio_put(folio);
+			goto release_sem;
+		}
+	}
+	jffs2_dbg(1, "end write_begin(). folio->flags %lx\n", folio->flags.f);
+
+release_sem:
+	mutex_unlock(&c->alloc_sem);
+out_err:
+	return ret;
+}
+
+static int jffs2_write_end(const struct kiocb *iocb,
+			   struct address_space *mapping,
+			   loff_t pos, unsigned len, unsigned copied,
+			   struct folio *folio, void *fsdata)
+{
+	/* Actually commit the write from the page cache page we're looking at.
+	 * For now, we write the full page out each time. It sucks, but it's simple
+	 */
+	struct inode *inode = mapping->host;
+	struct jffs2_inode_info *f = JFFS2_INODE_INFO(inode);
+	struct jffs2_sb_info *c = JFFS2_SB_INFO(inode->i_sb);
+	struct jffs2_raw_inode *ri;
+	unsigned start = pos & (PAGE_SIZE - 1);
+	unsigned end = start + copied;
+	unsigned aligned_start = start & ~3;
+	int ret = 0;
+	uint32_t writtenlen = 0;
+	void *buf;
+
+	jffs2_dbg(1, "%s(): ino #%llu, page at 0x%llx, range %d-%d, flags %lx\n",
+		  __func__, inode->i_ino, folio_pos(folio),
+		  start, end, folio->flags.f);
+
+	/* We need to avoid deadlock with page_cache_read() in
+	   jffs2_garbage_collect_pass(). So the folio must be
+	   up to date to prevent page_cache_read() from trying
+	   to re-lock it. */
+	BUG_ON(!folio_test_uptodate(folio));
+
+	if (end == PAGE_SIZE) {
+		/* When writing out the end of a page, write out the
+		   _whole_ page. This helps to reduce the number of
+		   nodes in files which have many short writes, like
+		   syslog files. */
+		aligned_start = 0;
+	}
+
+	ri = jffs2_alloc_raw_inode();
+
+	if (!ri) {
+		jffs2_dbg(1, "%s(): Allocation of raw inode failed\n",
+			  __func__);
+		folio_unlock(folio);
+		folio_put(folio);
+		return -ENOMEM;
+	}
+
+	/* Set the fields that the generic jffs2_write_inode_range() code can't find */
+	ri->ino = cpu_to_je32(inode->i_ino);
+	ri->mode = cpu_to_jemode(inode->i_mode);
+	ri->uid = cpu_to_je16(i_uid_read(inode));
+	ri->gid = cpu_to_je16(i_gid_read(inode));
+	ri->isize = cpu_to_je32((uint32_t)inode->i_size);
+	ri->atime = ri->ctime = ri->mtime = cpu_to_je32(JFFS2_NOW());
+
+	buf = kmap_local_folio(folio, aligned_start);
+	ret = jffs2_write_inode_range(c, f, ri, buf,
+				      folio_pos(folio) + aligned_start,
+				      end - aligned_start, &writtenlen);
+	kunmap_local(buf);
+
+	if (ret)
+		mapping_set_error(mapping, ret);
+
+	/* Adjust writtenlen for the padding we did, so we don't confuse our caller */
+	writtenlen -= min(writtenlen, (start - aligned_start));
+
+	if (writtenlen) {
+		if (inode->i_size < pos + writtenlen) {
+			inode->i_size = pos + writtenlen;
+			inode->i_blocks = (inode->i_size + 511) >> 9;
+
+			inode_set_mtime_to_ts(inode,
+					      inode_set_ctime_to_ts(inode, ITIME(je32_to_cpu(ri->ctime))));
+		}
+	}
+
+	jffs2_free_raw_inode(ri);
+
+	if (start+writtenlen < end) {
+		/* generic_file_write has written more to the page cache than we've
+		   actually written to the medium. Mark the page !Uptodate so that
+		   it gets reread */
+		jffs2_dbg(1, "%s(): Not all bytes written. Marking page !uptodate\n",
+			__func__);
+		folio_clear_uptodate(folio);
+	}
+
+	jffs2_dbg(1, "%s() returning %d\n",
+		  __func__, writtenlen > 0 ? writtenlen : ret);
+	folio_unlock(folio);
+	folio_put(folio);
+	return writtenlen > 0 ? writtenlen : ret;
+}