From: David Faust <david.faust@oracle.com>
To: "Jose E. Marchesi" <jose.marchesi@oracle.com>
Cc: binutils@sourceware.org, jbeulich@suse.com
Subject: Re: [PATCH v2] bpf: avoid creating wrong symbols while parsing
Date: Fri, 17 Nov 2023 12:09:36 -0800 [thread overview]
Message-ID: <6e00c2b6-a561-41dd-8701-e13627d9c57f@oracle.com> (raw)
In-Reply-To: <87edgo6uvr.fsf@oracle.com>
On 11/17/23 11:29, Jose E. Marchesi wrote:
>
> Hi David.
>
>> [ WAS: gas,bpf: cleanup bad symbols created while parsing
>> v1: https://sourceware.org/pipermail/binutils/2023-November/130556.html
>>
>> Changes from v1:
>> - Rewrite patch to avoid inserting wrong symbols in the first place,
>> rather than insert, remove, re-insert dance.
>> Suggested by Jan Beulich. ]
>>
>> To support the "pseudo-C" asm dialect in BPF, the BPF parser must often
>> attempt multiple different templates for a single instruction. In some
>> cases this can cause the parser to incorrectly parse part of the
>> instruction opcode as an expression, which leads to the creation of a
>> new undefined symbol.
>>
>> Once the parser recognizes the error, the expression is discarded and it
>> tries again with a new instruction template. However, symbols created
>> during the process are added to the symbol table and are not removed
>> even if the expression is discarded.
>>
>> This is a problem for BPF: generally the assembled object will be loaded
>> directly to the Linux kernel, without being linked. These erroneous
>> parser-created symbols are rejected by the kernel BPF loader, and the
>> entire object is refused.
>>
>> This patch remedies the issue by tentatively creating symbols while
>> parsing instruction operands, and storing them in a temporary list
>> rather than immediately inserting them into the symbol table. Later,
>> after the parser is sure that it has correctly parsed the instruction,
>> those symbols are committed to the real symbol table.
>>
>> This approach is modeled directly after Jan Beulich's patch for RISC-V:
>>
>> commit 7a29ee290307087e1749ce610207e93a15d0b78d
>> RISC-V: adjust logic to avoid register name symbols
>>
>> Many thanks to Jan for recognizing the problem as similar, and pointing
>> me to that patch.
>>
>> Tested on x86_64-linux-gnu host for bpf-unknown-none target.
>>
>> gas/
>>
>> * config/tc-bpf.c (parsing_insn_operands): New.
>> (parse_expression): Set it here.
>> (deferred_sym_rootP, deferred_sym_lastP): New.
>> (orphan_sym_rootP, orphan_sym_lastP): New.
>> (bpf_parse_name): New.
>> (parse_error): Clear deferred symbol list on error.
>> (md_assemble): Clear parsing_insn_operands. Commit deferred
>> symbols to symbol table on successful parse.
>> * config/tc-bpf.h (md_parse_name): Define to...
>> (bpf_parse_name): ...this. New prototype.
>> * testsuite/gas/bpf/asm-extra-sym-1.s: New test source.
>> * testsuite/gas/bpf/asm-extra-sym-1.d: New test.
>> * testsuite/gas/bpf/bpf.exp: Run new test.
>> ---
>> gas/config/tc-bpf.c | 92 +++++++++++++++++++++++++
>> gas/config/tc-bpf.h | 4 ++
>> gas/testsuite/gas/bpf/asm-extra-sym-1.d | 7 ++
>> gas/testsuite/gas/bpf/asm-extra-sym-1.s | 1 +
>> gas/testsuite/gas/bpf/bpf.exp | 3 +
>> 5 files changed, 107 insertions(+)
>> create mode 100644 gas/testsuite/gas/bpf/asm-extra-sym-1.d
>> create mode 100644 gas/testsuite/gas/bpf/asm-extra-sym-1.s
>>
>> diff --git a/gas/config/tc-bpf.c b/gas/config/tc-bpf.c
>> index fd4144a354b..3122f80804a 100644
>> --- a/gas/config/tc-bpf.c
>> +++ b/gas/config/tc-bpf.c
>> @@ -1223,6 +1223,7 @@ add_relaxed_insn (struct bpf_insn *insn, expressionS *exp)
>> See md_operand below to see how exp_parse_failed is used. */
>>
>> static int exp_parse_failed = 0;
>> +static bool parsing_insn_operands = false;
>>
>> static char *
>> parse_expression (char *s, expressionS *exp)
>> @@ -1230,6 +1231,9 @@ parse_expression (char *s, expressionS *exp)
>> char *saved_input_line_pointer = input_line_pointer;
>> char *saved_s = s;
>>
>> + /* Wake up bpf_parse_name before the call to expression (). */
>> + parsing_insn_operands = true;
>> +
>> exp_parse_failed = 0;
>> input_line_pointer = s;
>> expression (exp);
>> @@ -1251,6 +1255,71 @@ parse_expression (char *s, expressionS *exp)
>> return s;
>> }
>>
>> +/* Symbols created by this parse, but not yet committed to the real
>> + symbol table. */
>> +static symbolS *deferred_sym_rootP;
>> +static symbolS *deferred_sym_lastP;
>> +
>> +/* Symbols discarded by a previous parse. Symbols cannot easily be freed
>> + after creation, so try to recycle. */
>> +static symbolS *orphan_sym_rootP;
>> +static symbolS *orphan_sym_lastP;
>> +
>> +/* Implement md_parse_name hook. Handles any symbol found in an expression.
>> + This allows us to tentatively create symbols, before we know for sure
>> + whether the parser is using the correct template for an instruction.
>> + If we end up keeping the instruction, the deferred symbols are committed
>> + to the real symbol table. This approach is modeled after the riscv port. */
>> +
>> +bool
>> +bpf_parse_name (const char *name, expressionS *exp, enum expr_mode mode)
>> +{
>> + symbolS *sym;
>> +
>> + /* If we aren't currently parsing an instruction, don't do anything.
>> + This prevents tampering with operands to directives. */
>> + if (!parsing_insn_operands)
>> + return false;
>> +
>> + gas_assert (mode == expr_normal);
>> +
>> + if (symbol_find (name) != NULL)
>> + return false;
>
> Is this check correct?
>
> I see that expr.c:operand calls symbol_find_or_make right after calling
> md_parse_name (if the later is defined) and actually installs the
> resulting symbol in expressionP even if it was already defined. That is
> not done if md_parse_name returns `false'.
Hmm, I think you may have the logic in operand () flipped.
If md_parse_name returns 'false' then symbol_find_or_make is called;
if it returns 'true' then the 'break' is hit:
#ifdef md_parse_name
...
if (md_parse_name (name, expressionP, mode, &c))
{
restore_line_pointer (c);
break;
}
#endif
symbolP = symbol_find_or_make (name);
...
That is what we want to happen - if the symbol_find in parse_name
returns non-NULL, then we know a real symbol already exists for the
name we are looking at, and we should use it. In that case,
md_parse_name returns false, so operand () calls symbol_find_or_make,
which finds the extant symbol and installs it in the expression.
>
>> +
>> + for (sym = deferred_sym_rootP; sym; sym = symbol_next (sym))
>> + if (strcmp (name, S_GET_NAME (sym)) == 0)
>> + break;
>> +
>> + /* Tentatively create a symbol. */
>> + if (!sym)
>> + {
>> + /* See if we can reuse a symbol discarded by a previous parse.
>> + This may be quite common, for example when trying multiple templates
>> + for an instruction with the first reference to a valid symbol. */
>> + for (sym = orphan_sym_rootP; sym; sym = symbol_next (sym))
>> + if (strcmp (name, S_GET_NAME (sym)) == 0)
>> + {
>> + symbol_remove (sym, &orphan_sym_rootP, &orphan_sym_lastP);
>> + break;
>> + }
>> +
>> + if (!sym)
>> + sym = symbol_create (name, undefined_section, &zero_address_frag, 0);
>> +
>> + /* Add symbol to the deferred list. If we commit to the isntruction,
>> + then the symbol will be inserted into to the real symbol table at
>> + that point (in md_assemble). */
>> + symbol_append (sym, deferred_sym_lastP, &deferred_sym_rootP,
>> + &deferred_sym_lastP);
>> + }
>> +
>> + exp->X_op = O_symbol;
>> + exp->X_add_symbol = sym;
>> + exp->X_add_number = 0;
>> +
>> + return true;
>> +}
>> +
>> /* Parse a BPF register name and return the corresponding register
>> number. Return NULL in case of parse error, or a pointer to the
>> first character in S that is not part of the register name. */
>> @@ -1317,6 +1386,16 @@ parse_error (int length, const char *fmt, ...)
>> va_end (args);
>> partial_match_length = length;
>> }
>> +
>> + /* Discard deferred symbols from the failed parse. They may potentially
>> + be reused in the future from the orphan list. */
>> + while (deferred_sym_rootP)
>> + {
>> + symbolS *sym = deferred_sym_rootP;
>> + symbol_remove (sym, &deferred_sym_rootP, &deferred_sym_lastP);
>> + symbol_append (sym, orphan_sym_lastP, &orphan_sym_rootP,
>> + &orphan_sym_lastP);
>> + }
>> }
>>
>> /* Assemble a machine instruction in STR and emit the frags/bytes it
>> @@ -1606,6 +1685,10 @@ md_assemble (char *str ATTRIBUTE_UNUSED)
>> }
>> }
>>
>> + /* Mark that we are no longer parsing an instruction, bpf_parse_name does
>> + not interfere with symbols in e.g. assembler directives. */
>> + parsing_insn_operands = false;
>> +
>> if (opcode == NULL)
>> {
>> as_bad (_("unrecognized instruction `%s'"), str);
>> @@ -1622,6 +1705,15 @@ md_assemble (char *str ATTRIBUTE_UNUSED)
>>
>> #undef PARSE_ERROR
>>
>> + /* Commit any symbols created while parsing the instruction. */
>> + while (deferred_sym_rootP)
>> + {
>> + symbolS *sym = deferred_sym_rootP;
>> + symbol_remove (sym, &deferred_sym_rootP, &deferred_sym_lastP);
>> + symbol_append (sym, symbol_lastP, &symbol_rootP, &symbol_lastP);
>> + symbol_table_insert (sym);
>> + }
>> +
>> /* Generate the frags and fixups for the parsed instruction. */
>> if (do_relax && isa_spec >= BPF_V4 && insn.is_relaxable)
>> {
>> diff --git a/gas/config/tc-bpf.h b/gas/config/tc-bpf.h
>> index 9fb71eddd14..06096ef5926 100644
>> --- a/gas/config/tc-bpf.h
>> +++ b/gas/config/tc-bpf.h
>> @@ -51,6 +51,10 @@
>> a jump to offset 0 means jump to the next instruction. */
>> #define md_single_noop_insn "ja 0"
>>
>> +#define md_parse_name(name, exp, mode, c) \
>> + bpf_parse_name (name, exp, mode)
>> +bool bpf_parse_name (const char *, struct expressionS *, enum expr_mode);
>> +
>> #define TC_EQUAL_IN_INSN(c, s) bpf_tc_equal_in_insn ((c), (s))
>> extern bool bpf_tc_equal_in_insn (int, char *);
>>
>> diff --git a/gas/testsuite/gas/bpf/asm-extra-sym-1.d b/gas/testsuite/gas/bpf/asm-extra-sym-1.d
>> new file mode 100644
>> index 00000000000..113750dd3fd
>> --- /dev/null
>> +++ b/gas/testsuite/gas/bpf/asm-extra-sym-1.d
>> @@ -0,0 +1,7 @@
>> +#as: -EL -mdialect=pseudoc
>> +#nm: --numeric-sort
>> +#source: asm-extra-sym-1.s
>> +#name: BPF pseudoc no extra symbols 1
>> +
>> +# Note: there should be no output from nm.
>> +# Previously a bug in the BPF parser created an UND '*' symbol.
>> diff --git a/gas/testsuite/gas/bpf/asm-extra-sym-1.s b/gas/testsuite/gas/bpf/asm-extra-sym-1.s
>> new file mode 100644
>> index 00000000000..2cfa605a259
>> --- /dev/null
>> +++ b/gas/testsuite/gas/bpf/asm-extra-sym-1.s
>> @@ -0,0 +1 @@
>> + r2 = *(u32*)(r1 + 8)
>> diff --git a/gas/testsuite/gas/bpf/bpf.exp b/gas/testsuite/gas/bpf/bpf.exp
>> index 80f5a1dbc2d..fcbeccd8ecd 100644
>> --- a/gas/testsuite/gas/bpf/bpf.exp
>> +++ b/gas/testsuite/gas/bpf/bpf.exp
>> @@ -72,4 +72,7 @@ if {[istarget bpf*-*-*]} {
>> run_dump_test disp16-overflow-relax
>> run_dump_test disp32-overflow
>> run_dump_test imm32-overflow
>> +
>> + # Test that parser does not create undefined symbols
>> + run_dump_test asm-extra-sym-1
>> }
next prev parent reply other threads:[~2023-11-17 20:09 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-11-17 18:54 David Faust
2023-11-17 19:29 ` Jose E. Marchesi
2023-11-17 20:09 ` David Faust [this message]
2023-11-17 20:27 ` Jose E. Marchesi
2023-11-17 21:41 ` David Faust
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=6e00c2b6-a561-41dd-8701-e13627d9c57f@oracle.com \
--to=david.faust@oracle.com \
--cc=binutils@sourceware.org \
--cc=jbeulich@suse.com \
--cc=jose.marchesi@oracle.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).